refactor OCR engine configuration to use dependency pattern

[AdemBoukhris457]

Summary

Refactors OCR engine configuration in StructuredPDFParser and EnhancedPDFParser to use a dependency pattern. Instead of exposing individual OCR parameters directly in parser classes, users now initialize OCR engines externally and pass them as dependencies.

Motivation

• Clearer API: Users explicitly choose and configure one OCR backend at a time
• Prevent mixed configurations: Avoids confusion from mixing PyTesseract and PaddleOCR parameters
• Reusability: OCR engines can be created once and reused across multiple parsers
• Separation of concerns: OCR configuration is separated from parser logic

Changes

Code Changes

• Remove individual OCR parameters (ocr_lang, ocr_psm, ocr_oem, ocr_extra_config, paddleocr_*) from parser constructors
• Add single ocr_engine parameter accepting PytesseractOCREngine or PaddleOCREngine instances
• Maintain backward compatibility: creates default PytesseractOCREngine if ocr_engine=None
• Update CLI to instantiate OCR engines before passing to parsers
• Update UI components (Gradio apps) to create OCR engine instances
• Update all parser instantiation points across codebase

Documentation Changes

• Update README with new OCR engine examples
• Update all documentation files (user guides, API reference, examples)
• Add examples showing OCR engine reuse across parsers

Migration Guide

Before:

parser = StructuredPDFParser(
    ocr_lang="eng",
    ocr_psm=6,
    ocr_oem=3
)

After:

from doctra.engines.ocr import PytesseractOCREngine

tesseract_ocr = PytesseractOCREngine(lang="eng", psm=6, oem=3)
parser = StructuredPDFParser(ocr_engine=tesseract_ocr)

Default behavior (backward compatible):

parser = StructuredPDFParser()  # Still works, uses default PyTesseract