Add vision support for openrouter models

[few]

Aider v0.28.1-dev (9b2e6975843d253430c68ad011bcab27fb3e23cf) Model: openai/gpt-4-turbo using udiff edit format Git repo: none Repo-map: using 1024 tokens

openai/gpt-4-turbo (among other models) support vision, but aider says: "Cannot add image file as the model does not support image files"

Not sure how aider checks for vision support, in case it looks at https://openrouter.ai/api/v1/models, then checking $.architecture.modality=="multimodal" should be good enough.