Config for inline assistant model

[tallesborges]

I like to use a smart model on the assistant chat, but for inline assistant, this smart model is slow. So, I would like a config option, inline_model, that can use a smaller model like Codestral.