Add .docx parser for knowledge uploads

## Context
Dashboard and API accept PDF/TXT today; `.docx` was removed from file pickers until a parser exists.

## Scope
- Add `.docx` extraction (e.g. python-docx or unstructured)
- Wire into knowledge ingestion pipeline
- Unit test with a tiny fixture file
- Re-enable `.docx` in dashboard upload picker

## References
- `src/voxforge/modules/knowledge/`
- PM tasklist non-goal predecessor in Stage 1

Start in src/voxforge/modules/knowledge/ and trace the existing PDF/TXT ingestion path from the dashboard and API entry points. Add a tiny .docx fixture for unit coverage, verify extracted content reaches the knowledge pipeline, and confirm the dashboard upload picker accepts .docx.

Written by the indexing model from the issue text.