This adds support for a new way for specifying a "collection" of one
+item (that is, a + followed by an item id), which can of course contain
any number of files.  I believe the destination folder path is handled
well, including alphabucket support.  There may be some oddities yet.

Can't tell if these are coming from the archive or they were actually
uploaded, but it seems unlikely anyone would care about them.

It's the new way.

These files are never the ones we want.

This means if the last item in a download slot already exists, it
doesn't leave Verifying... on the screen forever.

Also makes it so 0 results is an error.  Also caps out parallel at the
number of items to download, in case of a very small result set.

Also check for successfully made error responses from the server.

Turns out sometimes the derivative files are bigger than the originals.

This involves scraping the HTML of the user's profile page.  The search
has to be done on their email address, not username, and I couldn't find
any other way to get it besides finding a video they've uploaded and
looking at the metadata.  There's got to be a better way!

Need to fix output...