Image-based writing prompts with the Unsplash API

It looks like the Unsplash API can return a random image from their high quality, free-use image library: https://unsplash.com/documentation#get-a-random-photo.

We should see if we can scope it down a bit from _all_ images (to get some targeted quality or let users pick between people, landscapes, objects, etc), but this seems like it'd be perfect for image-based writing prompts.

We don't even need to store answers in documents to begin with (#341); we can just show the image and expect users to write wherever they'd like, including their own documents if that's what they want.

Start with the Unsplash API documentation linked in the issue and review issue #341 for the constraint that answers need not be stored. Determine the smallest scope for targeted or selectable image prompts, then verify that the feature can show an image while letting users write wherever they choose.

Written by the indexing model from the issue text.