vision / image query support

[doctorguile]

So I added openai vision request implemented in my fork of org-ai

https://github.com/doctorguile/org-ai/blob/org/org-ai-block.el#L189

the problem is that currently messages is parsed before the model is determined

since I know what model (vision) I'm going to use before I write my prompt, that is not a problem (so I won't write a image query to 3.5 turbo for example)

but for others to use it, they need to aware of this pitfall.

I don't have enough itch to refactor that portion of org-ai since I don't know if it will be well received.

But if rksm or others have any ideas I would love to hear it

Thanks