merge from 3.3

#17403: urllib.parse.robotparser normalizes the urls before adding to
ruleline. This helps in handling certain types invalid urls in a conservative
manner. Patch contributed by Mher Movsisyan.