Feature specific model selection in LLM settings

[digitalWestie]

Allow user to choose model on a per-feature basis, starting with Sensemaker and Machine Learning.

Context:

I am thinking that eventually we are going to want this settings page amended to allow us to select different models for different applications. If using Vertex for example we might want to use Gemini-2.5.pro for sensemaking but something like flash for tagging as it is cheaper, has a higher quota and is faster?

Probably also want to move both sensemaker and machine learning feature switch in here as well

[digitalWestie]

Ideally we could allow admin to choose model before queuing a run

Choosing a model for task or feature probably means a back-end change in how we manage LLM Settings, probably a per feature override might be an idea, so we should see what core team thinks