[feature] add CLAP score metric to compute similarity between text and audio embeddings

[mcomunita]

• add CLAP score computation through CLAPScore class
• add test/test_clap_score.ipynb notebook to test new feature
• add test/clap_score_audio and test/clap_score_text with audio/text examples to run test

[gudgud96]

@mcomunita I fixed some merge conflicts in order to pass the unit tests.
Huge thanks for your contribution again! CLAP score is a super useful feature. Will do a new release soon.

[mcomunita]

Thanks @gudgud96. A useful extra would be to compute the CLAP score on the AudioCaps and the MusicCaps test sets and include the results in the README for reference. I haven't seen anyone reporting these numbers before, but I think it would give a useful lower bound.

[gudgud96]

That is a good idea. I am not working actively on AudioCaps / MusicCaps, but don't mind having a look.