Add token counting to rate limit

[ribenamaplesyrup]

OpenAI also sets limits on the number of tokens that can be passed to the API in a set period. Currently we're using asynciolimiter to limit requests by number of calls. Expanding this functionality to include token count would also be useful. Changes should be implemented in this notebook: https://github.com/Autonomy-Data-Unit/adulib/blob/main/nbs/api/llm.ipynb

[lukastk]

I had a quick look at this. I haven't implemented it just yet, as there's surprisingly no ready-made tool like asynciolimiter that scales well for this application.