Integrate bias model

[RomanPreuss]

I did an evaluation of the bias model with data generated using ChatGPT 3.5. The results are better compared to the click-bait results but still are not really good. Results can be found here:

https://github.com/TrustLevel/proof-of-concepts/tree/main/bias-detection/d4data

[RomanPreuss]

@dominiktilman the latest version is now also deployed to production and can be used using the local playground ui.

Open Points / Known Issues:

• [ ] If you submit a large text, the server runs into an Internal Error
• [ ] Currently, the endpoint is called /trustlevels and the response body returns a trustlevel score. After our last sync I understand that this actually should be contentQualityScore so we should rename it.
• [ ] The algorithm to compute the content quality score is quite simple and we should add thresholds indicating when and if a score is considered
• [ ] We should store the input, used weights, and resulting scores in a database. Furthermore, we should add a rating capability in the simple UI to vote if the resulting score is good or not good. This will allow us to potentially use this data later for validating the algorithm and potentially some trianing in the future.