feat: add support ollama api

[davidberenstein1957]

In this PR I've included basic support for ollama

Some design choices:

• use builtin urllib library. we might replace this with requests to also allow for passing a client
• retry 5 times with n+1 backoff in case of 500 (server errors) as fixed config for some robustness but to avoid spending a lot of time on this
• default to http://localhost:11434 but overwrite with OLLAMA_HOST env var as in Ollama docs
• generation kwargs don't align with other standards (maybe we can forcefully set them to deployment config when running ollama run model?)

Things to note:

• we use the api/chat endpoint
• we can't do n_generations in one request so we make multiple

[plaguss]

Nice addition!! For when you tackle the docs, it would be nice to include an example of how to setup ollama so it is as contained as possible

[davidberenstein1957]

Difficulty w.r.t. __init__ args is further covered here #246.

[plaguss]

Nice! Could you add some tests like the ones here? After that we should have a new cool integration 😃