We now also accept input from copied web pages and parse URLs from
anchors and images on the copied portion.

It's entirely possible for someone to copy-and-paste a whole chunk
of a document that includes both images and anchors. We probably
don't want to extract from the anchors if there are images. So we'll
first attempt to extract images, and fall back on anchors if there
are no images. If there are neither anchors nor images, we'll fall
back on generic parsing of plain text.

With the new URL parsing, it'd be possible that no URLs were parsed
at all. Let's provide some feedback for that.