Implement whisper.cpp

[mayeaux]

whisper.cpp is very performant: https://github.com/ggerganov/whisper.cpp

According to my testing, a $200 server with several TB of storage from OVH will perform a bit faster than equivalent Vast.ai servers which are somewhat unreliable and have limited storage space ($250 for 50GB storage).

Will need to refactor https://github.com/mayeaux/generate-subtitles/blob/master/transcribe/transcribing.js to wrap whisper.cpp , will also need a setup script to download the models and $ make the main file.

Progress reporting is implemented, frames/second is listed as a future enhancement: ggerganov/whisper.cpp#276