urldocument: add assert to make sure the url is normalized

This is later asserted again in the background parser, and somewhere
the assumption is violated (there are urldocuments with urls that contain
../ path segments). This assert might help find that.