Better support and docs for local models

[justyns]

Local models are currently supported as long as they expose an openai compatible api. It'd be better to have more documentation supporting this, along with some examples of installing/configuring them. This is pretty important for #34

Two local solutions I'd focus on right now are:

• https://github.com/ollama/ollama
• https://github.com/mudler/LocalAI

litellm doesn't run models itself, but would be good to document as a proxy to access other 3rd party APIs.

LocalAI seems like it's gaining a lot of new features, but ollama should be pretty simple for anyone to run locally (if they also run silverbullet locally)