Add CMU_ARCTIC dataset

[zakajd]

🚀 Feature

As for now, there is no such dataset in torchaudio.datasets.
More information about it can be found on official website: http://www.festvox.org/cmu_arctic/
List of all available speakers: http://www.festvox.org/cmu_arctic/packed/

Motivation

This dataset can be useful in some tasks (learning a voice embedding, voice cloning, TTS, STT).
It's rather small, so can be applicable for research purposes.

Pitch

from torchaudio.datasets import CMU_ARCTIC works.

If it's OK, I can work on this next week.

[vincentqb]

Thanks for suggesting this! Let's take a little bit of time to discuss choice of datasets that would be good to add: #550. Thoughts?

[vincentqb]

Since we do not yet have too many datasets, I'll say that, if you are willing to open a pull request implementing this dataset following our current templates, I'll be happy to review it. :) Would you like to do so?

[zakajd]

Yes, I'll do it on this weekend.