Allow prompt customization per interpreter model

[med8bra]

Describe the solution you'd like
Currently the prompt used with interpreter doesn't work well with all models especially small ones like llama3.2 (3b).

Adding an option to change the prompt as an advanced option will help user customize the interpreter feature per their model needs.

One issue maybe, since the interpreter expects key/value json object in model response which is mentioned in current prompt, customizing the prompt may break the feature, but I think it's worth it for power user. And these instructions could be specified as hints/rules around the prompt option.