Refactoring into API and GUI versions of ari implementation

[cansavvy]

See this powerpoint: https://docs.google.com/presentation/d/1Vjvq7PYuWsTkGi2EkXpnk0KtQYhbPSidBhMFQcqyb8I/edit

The idea is that mario will be the API version of a video maker while loqui will be the GUI version of a video maker. Both will be powered in the back end by functions here in ari.

Steps involved with this:

• Add free version of tts-coqui to these text2speech and ari functions
• Make GUI in loqui repo
• Update mario repo for deployment to Hutch servers
• Figure out billing for mario and loqui for when people want to use fancier tts options like Polly

[cansavvy]

For the GUI implementation and probably the API one, we would have a docker image that this would run on that already has https://github.com/coqui-ai/TTS#install-tts installed.