Anthropic instrumentation

[willbakst]

Closes #110

Some notes:

• Tried my best to refactor with OpenAI. This ultimately led to a decision to have a slightly worse internal DX due to lack of great typing for certain things. Wasn't sure how to resolve this without a dependency on any particular package in the new llm_providers.py file. If there are any suggestions for how to improve this, I'm all ears.
• I made some decisions with Anthropic to get it working with the UI, but I'm not certain this is the best behavior. I left a detailed comment in the code. Might be best to update the UI instead of doing what I did.

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

:loudspeaker: Thoughts on this report? Let us know!

[alexmojaki]

Thank you!!!