Support _multiple_ hosts (perhaps via --model)

[cori]

I run a few "local" models using a few different tools that are capable of presenting an OpenAI API-compatible interface. Sometimes I like to ask a few different models the same question to compare results. That works great if you're using a backend that accepts and respects the passed-in model name unless you want to run the same or similar queries against two different backends - let's say I want to compare the results from my local CodeLlama model to GPT-4's answer, or move between them to hone the final result.

It seems possible to support this by enhancing the "model" model - say to allow the value of the --model param to include an API url.

This would be useful to me but I'm not sure if it's too much of an edge case to balance the added complexity.