fix(docx): tolerate styles missing type

[jstar0]

Summary

Fixes the DOCX conversion crash from #2166 when a document contains a malformed word/styles.xml entry whose w:style element is missing the w:type attribute.

Root Cause / Context

MarkItDown pre-processes DOCX files and then hands them to Mammoth. Mammoth reads every style entry in word/styles.xml and indexes the w:type attribute directly. If a DOCX exporter writes a style without that attribute, Mammoth raises KeyError("w:type"), which MarkItDown reports as a FileConversionException.

Changes

• Remove malformed DOCX style entries that do not have w:type during the existing DOCX pre-processing step.
• Add a regression test that creates a DOCX fixture with one missing style type and verifies conversion still succeeds.

Scope / Risk

This only changes DOCX pre-processing for malformed style definitions. It does not alter document text, comments, equations, or valid style entries. If style cleanup itself fails, the pre-processor falls back to the original file content, matching the existing best-effort behavior used for math pre-processing.

Verification

cd packages/markitdown; hatch test tests/test_module_misc.py::test_docx_missing_style_type_does_not_crash
cd packages/markitdown; hatch test tests/test_module_misc.py::test_docx_comments tests/test_module_misc.py::test_docx_equations tests/test_module_misc.py::test_docx_missing_style_type_does_not_crash tests/test_module_vectors.py::test_convert_local -- -k docx
cd packages/markitdown; hatch test tests/test_module_misc.py
pre-commit run --files packages/markitdown/src/markitdown/converter_utils/docx/pre_process.py packages/markitdown/tests/test_module_misc.py

All commands passed.

Closes #2166