fix: handle DOCX files with inconsistent ZIP filename casing (#1812)

[lyydsheep]

Summary

Fixes #1812 — DOCX files with inconsistent ZIP filename casing between central directory and local file headers cause BadZipFile exceptions.

Some document generators (e.g. certain Microsoft Word versions, legal document systems) produce .docx files where the central directory records one casing (e.g. customXml/item2.xml) but the local file headers record another (e.g. customXML/item2.xml). Python's zipfile module raises BadZipFile when reading such files.

Root Cause

In pre_process_docx(), line 140 calls zip_input.read(name) for each file in the ZIP. Python's zipfile validates that the local file header filename matches the central directory entry at read time — a casing mismatch triggers:

BadZipFile: File name in directory 'customXml/item2.xml' and header b'customXML/item2.xml' differ.

Fix

Add _fix_zip_filename_casing() in pre_process.py that:

1. Reads the ZIP's central directory (authoritative source) to get correct filenames
2. Patches any local file header filenames that differ in casing
3. Only patches when lengths match (safe — only casing differs)
4. Returns a clean ZIP stream before any further processing

Called at the top of pre_process_docx(), so the fix applies to all downstream consumers (DocxConverter, DocxConverterWithOCR).

Test Plan

• New test test_docx_zip_filename_casing_mismatch — corrupts a valid .docx by uppercasing local header names, verifies zipfile.BadZipFile is raised without the fix, and confirms MarkItDown produces identical output with the fix
• Existing docx tests (test_docx_comments, test_docx_equations) still pass
• Full local test suite passes (10/10)

[lyydsheep]

@microsoft-github-policy-service agree