Support pasting images from clipboard into chat input

The chat input (<textarea>) currently only supports text. Users working with vision-capable models want to paste images from the clipboard alongside text.

ellmer already supports inline images via https://ellmer.tidyverse.org/reference/Content.html and chat$chat() accepts variadic content args, so the R-side integration is straightforward. 

Related: #172 (broader request for file/image input)

