mediawiki reader: improve strong/emph conformance

[silby]

cf. #10761 and #3044.

I made some progress with this today without completely blowing up the existing strong and emph parsers but weird edge cases remain. E.g. consider ''foo''''bar''. Pandoc today will give you Emph [ Str "foo" , Str "bar" ], which has an obvious appeal. My work in progress gives Emph [ Str "foo''" ] , Str "bar''", which is odder but defensible given other requirements for emphasized quote marks. The actual correct answer, according to MediaWiki, is Emph [ Str "foo'" , Strong [ Str "bar" ] ], i.e. foo'bar, which is basically a koan.

Parsoid has a lot of code just for processing quotes, presumably aiming to maintain bug-for-bug compatibility with whatever MediaWiki's first parser did. So what a string of single-quotes means varies depending on what comes after it in the line, in a more context-sensitive way than I expected.

Would it be better to merge code that makes us more conformant with MediaWiki for some cases and "wrong in a different way" for others, or to try to reach perfection?

[jgm]

In general I'm not too concerned with divergences in edge cases. Nobody is ever going to write ''foo''''bar'' and intend to get emph "foo'" + strong "bar". Your original case ''foo''', by contrast, seems like something that would come up naturally.

[jgm]

Is Parsoid the parser mediawiki uses? Or is that something else?