7db2832034
- waveform.py: fast ASCII waveform via ffmpeg PCM extract + numpy RMS
(avoids the slow per-frame astats approach)
- transcribe.py: faster-whisper integration with substring bad-word matching
- cli.py: new 'scan' (ASCII waveform + silence marks, no whisper)
and 'auto' (transcribe + bad-word flagging) subcommands
- numpy added to required dependencies
- 17 new tests (waveform + transcribe), 42 total all passing