DX-152: runpod-workers/worker-faster_whisper: update for modelcache

Haven't decided yet whether I'll fork this to a new repo entirely or just move this one over, since this ended up pretty much rewriting the whole thing from scratch.

• leverage the huggingface model cache and get rid of all the bits where we have to do very slow downloads at runtime.
• got rid of all the overabstracted nonsense, turning this into a handful of simple functions that run from the top to the bottom in a single file
• make it so that the tests can ever run
• actually comment the code so the flow isn't obscured
• update requirements.txt with latest the latest version of runpod
• try and give helpful error messages to the user!
• rewrite the entire dockerfile, etc.