[leilei3167]

Important

1. Contribution guidelines read. The change stays in the document extractor and its docx tests.
2. Open issues and pull requests searched. No open pull request already changes these files.
3. This pull request resolves the issue linked below.
4. Conventional Commits title used. This is not a breaking change.
5. The pull request title is suitable for the squash merge commit.
6. CLA signature is not included in this pull request.

Related Issue

Closes #346

Summary

Document Extractor fails on Strict Open XML saves and on docx packages that omit _rels/.rels. python-docx raises KeyError because the package relationship is not the transitional officeDocument type. Rewriting ZIP entry paths does not repair that, and word/document.xml is still present.

When opening the package fails, read word/document.xml directly and return the same paragraph and markdown-table text a normal docx already produces. Packages python-docx can open are unchanged. A zip with no word/document.xml still raises TextExtractionError.

Before the change, test_extract_text_from_docx_reads_package_when_opc_relationships_fail failed for a strict relationship type, a strict wordprocessingml document, and a package without _rels/.rels. Each failure was TextExtractionError wrapping KeyError for the missing officeDocument relationship. After the change, that test and the rest of tests/nodes/document_extractor pass (54). ruff, ty, and the import-layer check pass.

Checklist

• This pull request links the issue it resolves or advances
• This pull request title follows Conventional Commits, and any breaking change is marked with !
• If CLA Assistant prompted me, I signed CLA.md in the pull request conversation