improve data crawling

[razinskyy]

we need to think of a smarter way to collect tracks data, so that the process will be more accurate, relevant and faster. atm we thought of:

• minimizing genres options
• not adding tracks, if they are too distant from the track openai recommended