Rate limiting

[sd2k]

The reverse proxy should support rate limiting to enable Grafana admins to control costs due to expensive LLM calls. Things to consider:

• limits should be per provider, once we have a concept of provider (at the time of writing this would correspond to just one limit, for the the /openai path)
• over what time periods should we allow rate limiting? e.g. requests per minute, per hour, per day, etc.
• should we also limit API calls per client, for some definition of client? e.g. should user A have a separate limit to user B? should plugin X have a separate limit to plugin Y?

[sd2k]

Should be fairly easy using https://pkg.go.dev/go.uber.org/ratelimit#section-readme.