Ollama as content moderation provider

The problem here was that I had to map the responses to lookup tables for moderation models, since they are just chat models that answers in a specific way. The following models are available now:

https://ollama.com/library/shieldgemma
https://ollama.com/xe/llamaguard3

This means the provider needs to be updated when there comes a new moderation model.

Its been merged into dev release, anyone testing needs the dev version of AI since we did a minor update how the moderation endpoints works.