Allow Setting Model Per Task in Settings

[Routhinator]

Is your feature request related to a problem? Please describe.

Currently the Local AI plugin allows setting one default model, however my ollama instance has multiple models, and each has a specific use case they are best at. Image Generation and Description would preferably leverage the llava model rather than llama:3.2 - However this plugin currently does not allow per use case model settings. As a result, image requests fail unless I temporarily change the model setting to the correct one for image generation, and then change it back.

Describe the solution you'd like

Per task default models makes sense and would be ideal.

Describe alternatives you've considered

Not aware of any alternatives at the moment.

Additional context