feat: add local speaker diarization demo pipeline

[xyx-creator]

Summary

• add a local-only speaker diarization pipeline using pyannote
• add a standalone test.py entry for running speaker-aware transcription and sentiment locally
• keep the original Flask/API pipeline unchanged
• ignore the downloaded local diarization model directory in git

What Changed

• added app/models/speaker_diarization_model.py
• added app/data/speaker_diarization_data.py
• added app/services/speaker_diarization_service.py
• added app/services/local_audio_transcription_sentiment_pipeline.py
• added test.py
• updated .gitignore to ignore speaker-diarization-community-1/

Notes

• this PR is intentionally minimal-intrusion: it does not modify the existing Flask routes or the original audio_transcription_sentiment_pipeline
• the local demo pipeline uses a downloaded local pyannote model directory and assigns speaker labels such as speaker_1
• test.py also suppresses noisy third-party warnings so local output stays readable

Validation

• ran conda run -n saa python test.py
• ran conda run -n saa python -m pytest tests -q

Example Local Output

[1] 0.00s -> 3.00s | speaker_1 | NEU (0.8703)
[2] 3.00s -> 5.00s | speaker_1 | NEU (0.9020)