pdfnup was removed some time in 2021; pdfxup has always been around.

This is important in scenarios where AleKSIS, due to a race condition,
references something like a stale S3 content object. impressive-display
should log this and, most importantly, not try to handle the rror
output as display file.

File lists can now be used to define several sources to display.

This works close to the list feature of impressive, except that we do
not require an @ character, but instead treat all fiels ending in .txt
as list files.

Lists are scanned recursively, and fiels downloaded to generate the
new lists for impressive, and a final list is passed.

On download, all files are now compared to trigger a reload on change
of any file in the list.

Arguably, they should have been supported before, given that the
message after failure to analyze the file as a PDF outputs a WARNING,
so this is also a bugfix fixing that case. Previously, passing an
image file was perfectly ok, but it was then silently renamed to .pdf
and passed to a very confused impressive.