This makes the main download function return the path to the files it
just downloaded.  Might come in handy if I ever rewrite the release
script to not just parse the output from the download script.

This moves the files into subfolders by type, instead of author/title.
This should make it easier to omit certain download types from the
torrent, e.g. if you don't have a kindle and don't want any kindle
format books, but want the rest.

This adds a VERSION file containing the last updated time of the opds.

This refactors download into a module that can be included or invoked
from the command line.

The previous bash solution was brittle and had stopped working.  This
aims for feature parity in js, but working.  This should be easier to
expand in the future.

Note the release process is still in bash for now, as it was still doing
just fine.

Up to 2 retries, with delays between.

Let's keep download just as a download script, so it can be used by
anyone just wanting to do their own download.

The torrent creation bits should live in the release script instead.
This moves them there.  This also makes it so you can split up the wiki
page with a header, before <!-- EDITIONS FOLLOW -->, and the latest
edition will come right after that.

And tweak output.