Add support for Kyutai Streaming STT and TTS

Kyutai is an amazing set of Streaming Speech-To-Text and Text-To-Speech models that excel in English and French. It is permissively licensed CC-by-4.0 and has very strong performance.

Venice already has Text-To-Speech, which is quite good, but is missing streaming Speech-To-Text and it would be amazing to have a privacy-focused Speech-To-Text model.

Links:

https://kyutai.org/next/stt

https://kyutai.org/next/tts