Preserve PDF metadata when splitting chapters

[shankarpandala]

Motivation

• Splitting PDF chapters called set_metadata() twice which unintentionally overwrote the original document metadata (e.g. author, producer) when applying a per-chapter title.
• Add a regression test to ensure per-chapter titles are applied without losing existing metadata fields.

Description

• Replace the double set_metadata() calls in src/pdf_splitter/splitter.py with a single merged metadata dictionary so existing metadata is preserved while setting the chapter title.
• Construct a copy of the source metadata (dict(metadata)), set "title" to the chapter title, and pass that to new_doc.set_metadata().
• Add a new unit test tests/test_splitter.py that mocks fitz documents and verifies the split document receives merged metadata containing title, author, and producer.

Testing

• Ran PYTHONPATH=src pytest -q which executed the test suite including the new tests/test_splitter.py test.
• Test results: collected 5 items and the run completed with 4 passed, 1 skipped (the full test suite passed with the new regression test).
• The new test confirms the metadata merge behavior and the change prevents metadata loss when splitting.

---

Codex Task