feat: Implement rate limiting

[jonandernovella]

Please describe the feature

As a Mimir maintainer, I want to make sure that users' have reasonable limits over how many tokens they can request from ChatGPT.

Detailed description

• [ ] This limits could be enforced in different ways:
  • API key "K" can generate N tokens
  • User Account "U" can generate N tokens
  • IP address "A" can generate N tokens

The number of tokens "N" needs some research before we find a good number. AFAIK the maximum tokens per conversation is 4096. So if we want to allow a maximum 20 conversations per day, perhaps 20*4096 could be a good starting point?

To read more about rate limiting: https://platform.openai.com/docs/guides/rate-limits/overview

Acceptance criteria

• [ ] Clients who have used more than N tokens should be stopped from starting conversations with Mimir
• [ ] Integration tests should be implemented so that we make sure the limits are correctly in place

[jhagberg]

no fix