Multimoda file context support

[KorigamiK]

files such as images, pdfs, videos, audio should be allowed in the context for the models that support it.

changing to a model which doesn't have multimodal abilities should be a warning and the files stripped off but should still be allowed

[olimorris]

I plan on adding vision support (recommended reading is #1105) but no plans to add anything else for now.