Incorporate whisper.cpp for transcription

[vivekuppal]

Features

• Incorporate whisper.cpp as one of the potential offline STT services
• Incorporate the OAI latest large-v3 model
• Download models automatically as and when required

Primary benefit of incorporating this is that whisper.cpp STT is consistently 35-40% faster in transcription when using GPU as compared to OAI whisper module.

Usage:

python main.py -stt whisper.cpp

Fixes

• Move all log files to a separate logs folder
• Move all utilities to utilities folder
• Doc updates

Potential future improvements
Once below PR is merged it might remove the need for using ffmpeg to convert to 16 khz wav data file.
ggerganov/whisper.cpp#1549

Resolves #95