Chat: add vision support, upload images

[olafurpg]

Demo https://www.loom.com/share/638240e061cd449ab4fd2ac30b6ec356

Previously, it was only possible to send text to Cody via Chat. This PR adds new support to additionally send Cody images. This can be helpful, for example, when you want to write a basic HTML structure based on a Figma design.

Based on the changes in https://github.com/sourcegraph/cody/pull/4694/files#diff-93ab3b057cf286b165b92ef5a20dd588f286bf3d93ee78e72d63c4696904a006 In this PR, the image goes to Sourcegraph backend and works with both Anthropic and OpenAI models. The other PR only worked with the local llava model via Ollama.

Accompanying backend PR https://github.com/sourcegraph/sourcegraph/pull/546

Test plan

TBD

Changelog