Some SetName methods are make_setter!, some are make_atomic_setter!, and HTMLIFrameElement has its own DomRefCell reflecting its name. This inconsistency is a contributing factor in #25057, and cleaning it up is necessary to make a #25548 that won't crash.

I had changed the Element parent class to make names atoms in #25548, reflecting them in rare_data for lookup speed, but I had neglected to consider the differences between different Element subclasses there, and now that I look at all the specific ways different elements use names, a sweep to unify them all seems to be in order.