Requirements:

1. build read_audio function

• build module with read_audio function that reads .wav and .mp3
• address memory leak: extract audio without opening audio file, e.g. using base ffmpeg?

2. load and save to json instead of csv

• edit notebook code (that works with .csv data) to load json corpora instead
• save extracted audio as audio.json instead of csv column

3. delete ipynb before approving pull request