Assistant Panel: Image Input for Multi-modal Models

[ifsheldon]

I think Zed for now does not allow image input in the assistant chat panel. This is particularly useful for frontend development.

I guess the missing features are:

1. Display an image in Chat
2. Store a history that has images
3. Send images to the supported model services in base64
4. Settings for enabling image input
5. For those services that only accept image urls, add S3 configs for uploading images