Speaker diarization with whisper transcription

[Jiltseb]

This PR addresses the feature request #176.

1. Combines output from Speaker Diarization Deployment and Whisper deployment.
2. The post-processing algorithm generates diarized transcription as shown in the example notebook.
3. Added test for the post-processing algorithm.
4. Changes in docs

[movchan74]

I cannot comment the notebook, so I will write the comments here.

1. You can use the following code instead of your data filtering code:

[s.model_dump(include=["text", "time_interval", "speaker"]) for s in segments]

2. I cannot run the code because I get the following error:

RuntimeError: Traceback (most recent call last):
  File "/workspaces/aana_sdk/aana/deployments/pyannote_speaker_diarization_deployment.py", line 77, in apply_config
    self.diarize_model = Pipeline.from_pretrained(self.model_id)
  File "/root/.cache/pypoetry/virtualenvs/aana-vIr3-B0u-py3.10/lib/python3.10/site-packages/pyannote/audio/core/pipeline.py", line 138, in from_pretrained
    pipeline = Klass(**params)
  File "/root/.cache/pypoetry/virtualenvs/aana-vIr3-B0u-py3.10/lib/python3.10/site-packages/pyannote/audio/pipelines/speaker_diarization.py", line 130, in __init__
    model: Model = get_model(segmentation, use_auth_token=use_auth_token)
  File "/root/.cache/pypoetry/virtualenvs/aana-vIr3-B0u-py3.10/lib/python3.10/site-packages/pyannote/audio/pipelines/utils/getter.py", line 89, in get_model
    model.eval()
AttributeError: 'NoneType' object has no attribute 'eval'

I did set the HF_TOKEN environment variable to my Hugging Face token. And the model page says: You have been granted access to this model. So, I don't know what is the problem.

[Jiltseb]

I have changed the post-processing function into a class to keep the logic organized and the module-specific functions private. However, it has not been moved to diarization deployment because it will kill the flexibility to run the diarization and whisper models in parallel if needed. I think we can keep it this way until we have a specific SDK component that can combine the deployments and additional functions.

[movchan74]

Looks good! Great work 👍