Refactoring of Coqui TTS extension

[Tony-sama]

Simplified coqui extension by reducing to the following functionalities:

• check if a model installed
  • just look for files
  • check if install is corrupt (fail download)
• install a model
  • using coqui api automatic download, can be used to reinstall a model
• generate tts
  • much simpler than before
  • assume request payload provide proper model id and options

Notes:

• Only tested on gpu, but should work on cpu too.
• Can be optimized:
  • by keeping model loaded (maybe already managed by coqui api)
  • by keeping multiple model loaded for group chat (not sure the api allows it)