Add .docx support to RAG document loader

[pcbeingused333]

Closes #97 (or: relates to #97 — see note below)

What

Adds Word (.docx) support to fenn/agents/rag/loader.py, following the same
pattern as the existing _read_pdf loader.

• .docx added to SUPPORTED_EXTENSIONS
• _read_file now routes .docx files to a new _read_docx helper
• _read_docx extracts text from both paragraphs and table cells using python-docx
• New optional extra rag-docx = ["python-docx>=1.0.0"] (also added to rag-all)
• Unit tests in tests/unit/agents/test_docx_loader.py

How it was tested

pytest tests/unit/agents/test_docx_loader.py — 6 passing:
paragraphs, table cells, empty document → None, _read_file delegation,
load_documents, and the missing-dependency ImportError.

Note / question for maintainers

The existing loader error messages reference cofone[pdf] / cofone[web], while
the package is fenn and the extras are rag-pdf / rag-web. I kept the same
wording (cofone[docx]) in the new _read_docx for consistency with the current
file — but happy to align everything to fenn[rag-docx] in this PR (or a separate
one) if you prefer. Just let me know which you'd rather have.

[ApusBerliozi]

Hey @pcbeingused333! Everything looks good for me.

@blkdmr, could you please take a look at this PR as well?