Feature Request: Handle thinking models better

If you load a thinking model, it'll just output the thinking text like everything else. Most GUIs allow this to be either collapsed or hidden. I think hiding it would be just fine for the purposes of a clippy desktop assistant.

As a little bonus, it would be great if Clippy just looped the scratching his head animation while in the thinking process, and maybe just use that animation only when he's thinking so it'd be a clear way to signify what clippy is doing in that moment.

No files or tests are named in the issue. Start by locating the desktop assistant's model-output handling and Clippy animation handling; done means thinking text is hidden and the scratching-head animation loops only while the model is thinking.

Written by the indexing model from the issue text.