Previously, image extraction wouldn't catch images with a query string
(or anything else) appended. This fixes that by parsing extracted URLs
and only checking the path for what looks like an image file.