i tried adding a 'loopback' mode to the specialremote itself, but that
turned out to be too error-prone to really work, so for now running
tests requires an sshd.

this results in a semi-workable thing. the biggest problem is that
special remotes aren't really supposed to know much (or anything) about
the files they store, which is an assumption I break in the most direct
way possible: it only makes sense to store pdfs on a reMarkable.

So this now uses exiftool to get the pdf's title. Unfortunately, most of
my pdfs turn out to not have any titles – and those that have often have
useless titles set (e.g. "Microsoft Word Document" or some such).

Will have to think about that one a bit ..

most things work, but xochitl wants more meta data before things'll work