Working caption display before timing improvements

Gemini-suggested fix: Install server deps from audio-trimmer-server d…

…irectory

Progressive captions + performance logging

Caption fixes:
- Word-by-word progressive display (matches mobile app)
- Shows last 8 words spoken so far
- Fixes utterance-only display issue

Performance improvements:
- Frame render timing logs
- Slow frame warnings (>2s)
- Caption processing monitoring
- Progress milestones every 30 frames

Known issues:
- Fontconfig errors may slow rendering
- No font configuration fixes included

Working video generation without captions

- Font optimization: reduced from 9,000+ to 42 fonts
- Fixed fontconfig initialization for proper font rendering
- Video generation completes in ~3 minutes without hanging
- Roboto and DejaVu fonts properly configured
- Ready for caption integration

STABLE: Audio2 captions fully working

- Uses audioCodec('copy') to avoid transcoding
- Confirmed working on Railway production
- Last known good state before server-side video development