Port pocket-tts to supertonic3: 44100 Hz, 10 voices, 31 languages

Runtime model switching, per-model voice caching, thread-safety fix (#12)

add proper opus support

rename voices, implement better UI to search voices (#4)

Revert "initial attempt at supporting speed parameter"

initial attempt at supporting speed parameter

hide "absolute path" voice upload in Docker

add API docs to UI

update to pocket-tts 1.03

Initial commit