Add access token parameter for pyannote models

[juanmc2005]

Problem

The new release of pyannote.audio comes with new authentication requirements to download models, so diart users using pyannote models will need to go through this authentication procedure, but diart does not provide a way to download huggingface models with an access token.

Idea

Add a parameter hf_access_token to PipelineConfig and to any pyannote model constructors.

Example

from diart import PipelineConfig, OnlineSpeakerDiarization

config = PipelineConfig(hf_access_token="...")
diarization = OnlineSpeakerDiarization(config)

# or

from diart.models import SegmentationModel, EmbeddingModel

segmentation = SegmentationModel.from_pyannote("pyannote/segmentation", hf_access_token="...")
embedding = EmbeddingModel.from_pyannote("pyannote/embedding", hf_access_token="...")
config = PipelineConfig(segmentation=segmentation, embedding=embedding)
diarization = OnlineSpeakerDiarization(config)

# or

from diart import settings

settings.set_global_hf_access_token("...")
diarization = OnlineSpeakerDiarization()

[juanmc2005]

Implemented in #108