added faster whisper pipeline

[ben256]

No description provided.

[ben256]

For reference about changes: (whisperx/asr.py)

"""
Essentially added additional inference method which uses the faster-whisper batched inference pipeline instead of
WhisperX pipeline which doesn't work for arabic-english code switching.
For use with experimental version of faster-whisper SYSTRAN/faster-whisper#936
Run "pip install git+https://github.com/MahmoudAshraf97/faster-whisper.git@same_vad" to install
This version doesn't change the WhisperModel class which is used by whisperX for the legacy inference method, but does
provide the faster-whisper.BatchedInferencePipeline so for now it's our best bet.
This has been designed such that minial changes are required to the transcription engine as both pipelines use the same
function names.
"""