Refactor codebase into classes

[kouloumos]

All the logic has been moved and slightly refactored into separate
classes to remove duplicate code, scattered logic and achieve better
readability and maintanability of the codebase by having a clear flow
of the process.

An overview of the classes:

• Transcription is the main class that contains Transcripts.
• Each Transcript holds a Source which we want to transcribe,
  and it is either Audio, Video or Playlist.

How the flow looks like:

• We initialize a Transcription object that holds all the
  related configurations for the current transcription process
• We can add as many sources as we want to the current transcription
  with transcription.add_transcription_source(...)
• when we are ready, we transcription.start(), which:
  • produces an audio file by processing the source. This step is
    responsible for any downloads or conversions that needs to happen.
  • produces the transcription by processing the audio file. This step
    includes any summarizations, chapter generations, diarization that we
    might have configure.
    It can optionally:
    • write the transcription to a markdown file.
    • open a PR to the repo.
    • upload the transcription to an AWS S3 Bucket.
    • push the transcript to a Queuer backend.
    • write the payload of the transcription to a json file.

Also, updated the tests based on the new flow of the refactored codebase.
The Deepgram related methods that are still in application.py can later
become their own class as we start adding more services.