Add support for Ollama backend

[evangineer]

For those of us running our LLMs locally, Ollama has been a boon and it provides an OpenAI compatible chat completion API, so it's not too difficult to add support if you are already supporting OpenAI.

I might implement this myself if I can find the spare time.