Followup to
r86061: add parser test case to confirm that '&apos' in wikitext doesn't make it to output.
''' is not defined in HTML 4, and some browsers (for instance IE at least through 8) do not support it.
It is however part of XML 1 (and thus XHTML 1), and is also defined in HTML 5 provisional, so it's fine to accept it.
Normalization seems to already handle turning the ' named char reference into ' so we don't have any compatibility worries.
Also added a comment on the def in Sanitizer that it shouldn't be used in output.