Adds custom agents to the langchain benchmarking repo

[eyurtsev]

• This PR adds code for running custom agents to the langchain benchmarking repo.
• The agent code is good enough for experimentation / prototyping, but I don't think it's good enough for the langchain repo:
  -- The abstractions aren't fully implemented and aren't ready for production use -- but OK for research
  -- For production use, one may want to remove all the intermediate abstractions to keep the agent as simple as possible

I was thinking initially of including this in a different repo, but I think it's over-complicating things, probably OK to include some reference implementations inside of langchain benchmarks.

[eyurtsev]

Continuing work from here: #109

But some changes have been merged (e.g., model registry) and benchmarking code etc

[hinthornw]

Heh got caught mid review but looks good!