Add Audio Feature Extraction Task

[ecyht2]

Currently, the Feature Extraction task includes both models for audio and text feature extraction (it is officially placed under the NLP modality). I think it would be nice to have a new task for Audio Feature Extraction task just like Image Feature Extraction to better label models.

Some Audio Feature Extraction models in Feature Extractions:

• facebook/encodec
• kyutai/mimi

[osanseviero]

Yes, having an audio-feature-extraction has been discussed few times in the past and there's agreement having it could make sense. cc @Vaibhavs10 @pcuenca