fix(docx): ignore malformed styles missing type

[gingerninja85]

Summary

Fixes a DOCX conversion crash when word/styles.xml contains a malformed w:style entry without the required w:type attribute.

Mammoth indexes style_element.attributes["w:type"], so these malformed style records currently bubble up as:

DocxConverter threw KeyError with message: 'w:type'

This patch extends DOCX preprocessing to remove only malformed style definitions missing w:type or w:styleId, allowing body text conversion to continue instead of failing the whole document.

Fixes #2166.

Verification

pytest packages/markitdown/tests/test_module_vectors.py::test_convert_docx_with_style_missing_type -q
# 1 passed

pytest 'packages/markitdown/tests/test_module_vectors.py::test_convert_local[test_vector0]' \
  'packages/markitdown/tests/test_module_vectors.py::test_convert_stream_with_hints[test_vector0]' \
  'packages/markitdown/tests/test_module_vectors.py::test_convert_stream_without_hints[test_vector0]' \
  'packages/markitdown/tests/test_module_vectors.py::test_convert_file_uri[test_vector0]' \
  'packages/markitdown/tests/test_module_vectors.py::test_convert_data_uri[test_vector0]' \
  packages/markitdown/tests/test_module_vectors.py::test_convert_docx_with_style_missing_type -q
# 6 passed

pre-commit run --files packages/markitdown/src/markitdown/converter_utils/docx/pre_process.py packages/markitdown/tests/test_module_vectors.py
# black Passed

python -m compileall -q packages/markitdown/src packages/markitdown/tests
# passed

I also ran hatch test; it reached 332 passed, 4 skipped with one local environment failure unrelated to this change because ffprobe is not installed for tests/test_module_misc.py::test_speech_transcription.