feat: add ground truth for benchmark quality scoring across all non-archive file types

Add ~469 ground truth text files covering all non-archive, non-image file types,
expanding quality score coverage from 77 to 546 fixtures (95% of all benchmarks).

Ground truth generated using independent tools not part of benchmarked frameworks:
- Raw source copy for text formats (md, txt, rst, org, tex, toml, yaml, etc.)
- pdftotext for PDFs, python-docx/pptx/openpyxl for Office XML formats
- pandoc for epub, fb2, docbook, odt, rtf, opml
- beautifulsoup for HTML, Python email stdlib for EML, extract-msg for MSG
- odfpy for ODS, xlrd for XLS, nbformat for ipynb, textutil for legacy doc/ppt

Includes generation script, fixture JSON patches, and updated ground truth mapping.

Author: Goldziher

test_documents/ground_truth/bib/bib_comprehensive.txt

@@ -0,0 +1,151 @@
+@article{entry_article_one,
+  author = {Alice Smith and Bob Johnson},
+  title = {Exploring Article Formats},
+  journal = {Journal of Examples},
+  year = {2012},
+  volume = {42},
+  number = {3},
+  pages = {101-120}
+}
+
+@book{entry_book_foundations,
+  author = {Carla Ruiz and Deepak Patel},
+  title = {Foundations of Structured Documents},
+  publisher = {Example Press},
+  year = {2010},
+  address = {Berlin}
+}
+
+@inproceedings{entry_inproceedings_parallel,
+  author = {Ethan Lee and Fatima Khan},
+  title = {Parallel Parsing Techniques},
+  booktitle = {Proceedings of the Structured Data Conference},
+  year = {2019},
+  pages = {55-68}
+}
+
+@phdthesis{entry_phdthesis_semantics,
+  author = {Grace Muller},
+  title = {Semantics Aware Document Pipelines},
+  school = {University of Hamburg},
+  year = {2015}
+}
+
+@mastersthesis{entry_mastersthesis_design,
+  author = {Hugo Silva},
+  title = {Designing Resilient Extraction Systems},
+  school = {Technical University of Lisbon},
+  year = {2018}
+}
+
+@techreport{entry_techreport_scalability,
+  author = {Ingrid Novak},
+  title = {Scalability Benchmarks for Extraction Engines},
+  institution = {Alpine Research Labs},
+  number = {ARL-TR-011},
+  year = {2011}
+}
+
+@manual{entry_manual_reference,
+  author = {Javier Torres},
+  title = {Kreuzberg Reference Manual},
+  organization = {Kreuzberg Labs},
+  year = {2005}
+}
+
+@misc{entry_misc_dataset,
+  author = {Keiko Tanaka},
+  title = {Annotated Extraction Dataset},
+  year = {2022},
+  howpublished = {Dataset Repository},
+  note = {Version 3.2}
+}
+
+@unpublished{entry_unpublished_notes,
+  author = {Liam O'Connor},
+  title = {Notes on Incremental Extraction},
+  note = {Draft manuscript},
+  year = {2021}
+}
+
+@incollection{entry_incollection_story,
+  author = {Mei Huang},
+  title = {Story Driven Testing},
+  booktitle = {Modern Document Pipelines},
+  publisher = {Insight Publishing},
+  year = {2016},
+  chapter = {4}
+}
+
+@inbook{entry_inbook_chapter,
+  author = {Noah Becker},
+  title = {Advanced Pipeline Patterns},
+  chapter = {11},
+  publisher = {TechWorks},
+  year = {2009}
+}
+
+@proceedings{entry_proceedings_ai,
+  title = {Proceedings of the AI Extraction Summit},
+  year = {2017},
+  editor = {Olivia Rossi}
+}
+
+@booklet{entry_booklet_summary,
+  author = {Priya Desai},
+  title = {Summary of Extraction Benchmarks},
+  year = {2014},
+  howpublished = {Internal Memo}
+}
+
+@article{entry_article_modern,
+  author = {Quinn Parker},
+  title = {Modern Approaches to Layout Analysis},
+  journal = {International Journal of OCR},
+  year = {2020},
+  volume = {7},
+  number = {1}
+}
+
+@book{entry_book_distributed,
+  author = {Rina Haddad},
+  title = {Distributed Text Processing},
+  publisher = {Northern Lights},
+  year = {2002}
+}
+
+@inproceedings{entry_inproceedings_reproducibility,
+  author = {Samuel Ortega},
+  title = {Reproducible Extraction Pipelines},
+  booktitle = {Workshop on Reliable NLP},
+  year = {2013},
+  pages = {12-20}
+}
+
+@article{entry_article_architecture,
+  author = {Talia Cohen},
+  title = {Architecture Patterns for Extractors},
+  journal = {Systems Journal},
+  year = {2008}
+}
+
+@techreport{entry_techreport_innovation,
+  author = {Umar Farouk},
+  title = {Innovation in Document Processing},
+  institution = {Global Research Institute},
+  year = {2006}
+}
+
+@booklet{entry_booklet_research,
+  author = {Valeria Costa},
+  title = {Research Highlights in OCR},
+  year = {1998},
+  note = {Conference supplement}
+}
+
+@article{entry_article_future,
+  author = {Wei Zhang},
+  title = {Future of Unified Extraction},
+  journal = {Journal of Pipeline Research},
+  year = {2024}
+}

test_documents/ground_truth/commonmark/commonmark_sample.txt

@@ -0,0 +1,136 @@
+# CommonMark Sample Document
+
+This is a comprehensive CommonMark document showcasing standard markdown elements for testing and benchmarking purposes.
+
+## Introduction
+
+CommonMark is a **standardized specification** of Markdown. This document demonstrates *all* the core elements that should be properly parsed and extracted.
+
+### About Markdown
+
+Markdown is a lightweight markup language with plain-text formatting syntax. It's designed to be *easy to read* and **easy to write**.
+
+## Text Formatting
+
+You can use various text formatting options:
+
+- **Bold text** using double asterisks or underscores
+- *Italic text* using single asterisks or underscores
+- ***Bold and italic*** combined
+- `inline code` for short snippets
+
+## Lists
+
+### Unordered Lists
+
+- First item
+- Second item
+  - Nested item 1.1
+  - Nested item 1.2
+- Third item
+  - Nested item 3.1
+    - Double nested 3.1.1
+    - Double nested 3.1.2
+
+### Ordered Lists
+
+1. First step
+2. Second step
+   1. Sub-step 2.1
+   2. Sub-step 2.2
+3. Third step
+   1. Sub-step 3.1
+4. Fourth step
+
+## Code Blocks
+
+### Inline Code
+
+Use the `function_name()` method to process data.
+
+### Code Blocks
+
+```rust
+fn hello_world() {
+    println!("Hello, World!");
+}
+```
+
+```python
+def fibonacci(n):
+    if n <= 1:
+        return n
+    return fibonacci(n-1) + fibonacci(n-2)
+```
+
+```javascript
+const greeting = (name) => {
+    return `Hello, ${name}!`;
+};
+```
+
+## Links and References
+
+Visit the [CommonMark specification](https://spec.commonmark.org/) for complete documentation.
+
+You can also use [reference-style links][commonmark-spec] if preferred.
+
+[commonmark-spec]: https://spec.commonmark.org/
+
+## Blockquotes
+
+> This is a blockquote. It's useful for highlighting important information or quotes from other sources.
+
+> Blockquotes can contain **bold text** and *italic text*.
+>
+> They can also span multiple paragraphs.
+
+## Horizontal Rule
+
+---
+
+## Tables
+
+| Language | Type | Package Manager |
+|----------|------|-----------------|
+| Python | Dynamic | pip |
+| Rust | Compiled | cargo |
+| JavaScript | Dynamic | npm |
+| Java | Compiled | Maven |
+
+## Images
+
+![Alt text for image](https://example.com/image.png)
+
+## Mixed Content
+
+Here's a paragraph with multiple types of formatting. It contains **bold**, *italic*, `code`, and [links](https://example.com).
+
+### Complex Nested List
+
+1. First item with **bold** and *italic*
+   - Sub-item with `code`
+   - Another sub-item with [link](https://example.com)
+2. Second item
+   - Multiple levels
+     - Level 3 item 1
+     - Level 3 item 2 with **formatting**
+   - Back to level 2
+3. Third item
+
+## Conclusion
+
+This CommonMark document includes all standard elements:
+- Headers at multiple levels
+- Paragraphs with inline formatting
+- Unordered and ordered lists
+- Nested lists
+- Code blocks with syntax highlighting
+- Inline code
+- Links and reference-style links
+- Blockquotes
+- Horizontal rules
+- Tables
+- Mixed formatting and content
+
+The document is structured for comprehensive testing of markdown parsing and extraction capabilities.

test_documents/ground_truth/csv/csv_simple.txt

@@ -0,0 +1,5 @@
+Stanley Cups,,
+Team,Location,Stanley Cups
+Blues,STL,1
+Flyers,PHI,2
+Maple Leafs,TOR,13

test_documents/ground_truth/csv/csv_tables.txt

@@ -0,0 +1,11 @@
+Name,Department,Salary,Start Date,Active
+Alice Johnson,Engineering,95000,2020-01-15,true
+Bob Smith,Marketing,75000,2019-06-01,true
+Carol White,Sales,82000,2021-03-10,false
+David Brown,Engineering,105000,2018-11-20,true
+Emily Chen,HR,70000,2022-02-28,true
+Frank Davis,Operations,88000,2020-07-15,false
+Grace Lee,Marketing,79000,2021-09-01,true
+Henry Wilson,Engineering,115000,2017-04-12,true
+Isabella Martinez,Sales,91000,2019-12-05,true
+Jack Thompson,HR,68000,2023-01-10,true

test_documents/ground_truth/csv/test_mskanji.txt

@@ -0,0 +1,4 @@
+–¼‘O,”N—î,ZŠ
+²“¡‘¾˜Y,30,“Œ‹ž
+ŽO–؉pŽq,25,‘åã
+îà‹´~,35,–¼ŒÃ‰®

test_documents/ground_truth/djot/djot_tables.txt

@@ -0,0 +1,56 @@
+Simple table with caption:
+
+| Right | Left | Center | Default |
+|------:|:-----|:------:|-------|
+|    12 | 12   |   12   | 12      |
+|   123 | 123  |  123   | 123     |
+|     1 | 1    |   1    | 1       |
+
+^ Demonstration of simple table syntax.
+
+Simple table without caption:
+
+| Right | Left | Center | Default |
+|------:|:-----|:------:|-------|
+|    12 | 12   |   12   | 12      |
+|   123 | 123  |  123   | 123     |
+|     1 | 1    |   1    | 1       |
+
+Simple table indented two spaces:
+
+| Right | Left | Center | Default |
+|------:|:-----|:------:|-------|
+|    12 | 12   |   12   | 12      |
+|   123 | 123  |  123   | 123     |
+|     1 | 1    |   1    | 1       |
+
+^ Demonstration of simple table syntax.
+
+Multiline table with caption:
+
+| Centered Header | Left Aligned | Right Aligned | Default aligned                                       |
+|:---------------:|:-------------|--------------:|:------------------------------------------------------|
+|      First      | row          |          12.0 | Example of a row that spans multiple lines.           |
+|     Second      | row          |           5.0 | Here's another one. Note the blank line between rows. |
+
+^ Here's the caption. It may span multiple lines.
+
+Multiline table without caption:
+
+| Centered Header | Left Aligned | Right Aligned | Default aligned                                       |
+|:---------------:|:-------------|--------------:|:------------------------------------------------------|
+|      First      | row          |          12.0 | Example of a row that spans multiple lines.           |
+|     Second      | row          |           5.0 | Here's another one. Note the blank line between rows. |
+
+Table without column headers:
+
+|----:|:----|:---:|----:|
+|  12 | 12  | 12  |  12 |
+| 123 | 123 | 123 | 123 |
+|   1 | 1   |  1  |   1 |
+
+Multiline table without column headers:
+
+|:------:|:----|-----:|-----------------------------------------------------|
+| First  | row | 12.0 | Example of a row that spans multiple lines.           |
+| Second | row |  5.0 | Here's another one. Note the blank line between rows. |