Improve unit test generation

[MrSerth]

The purpose of this ticket is to collect ideas for improving the unit test generation process.

• per-user tokens for ChatGPT (Allow users to configure a personal OpenAI API key #1496)
• showing the model's error message to users
• A more visible box for the generate unit test button
• improving the file name (extension + name)
• Thinking of further "promotions"
• Maybe have this view as part of the "edit" view (since this could allow edits before they are persisted). However, this also "hides" the feature more which is not good
• Check a need for rate limiting
• Improved "billing" overview (i.e., on a per-user base)
• Testing the current implementation as part of a user study
• Improved model output as a json. Could help improve data extraction from the model (OpenAI: Use Structured Outputs #1627)

Probably, we should create sub-tickets for the respective items before tackling them.

[MrSerth]

Since we switched to a per-user API key, there is no need for billing or rate limiting any longer (users are spending their own money).