em and strong fail if html with _ inside

[Cerdic]

This test case works fine
___em strong <a href="http://parsedown.org/" class='externallink'>Parsedown</a>___

Whereas this fails
___em strong <a href="http://parsedown.org/" class='external_link'>Parsedown</a>___

[wkpark]

Similar case

**em <a href="#" alt="*">*em*</a>**

Pasedown does not handle these cases correctly.

[wkpark]

Another case

**[**FOO**](/bar/)**

FOO GFM also fail to parse it.

http://johnmacfarlane.net/babelmark2/?text=**%5B**FOO**%5D(%2Furl%2F)**

[Cerdic]

This last case is a bit different as strong in strong is not supposed to be supported in any case.
For the **em <a href="#" alt="*">*em*</a>**case I thing we could fix it in a similar way of first case
Cerdic@5caabe0

[wkpark]

@Cerdic all examples are correctly rendered by the Markdown 1.0.2b8

[Cerdic]

@wkpark ok my bad !

[aidantwoods]

Closing for the reasons outlined in #588

[Cerdic]

Too bad :(
I had to patch last version again in Cerdic@870839b as the use-case I initialy described is quite frequent in the CMS I'm using where URL in plain text are auto-changed into link with a class containing underscore char :
___em strong <a href="http://parsedown.org/" class='external_link'>Parsedown</a>___

I understand that the patch doesn't fix all cases, especialy the one described in #588 but I think that ignoring html tags in the regexp as proposed gives a much better robustness for a not a such perf impact.

[aidantwoods]

What happens if you try something like ___em strong <a href="http://parsedown.org/" class='external_link'>foo_bar</a>___ with the patched regex?

[Cerdic]

it fails :(

[aidantwoods]

If there were an easy fix for most cases then I'd definitely be up for getting Parsedown to behave a bit better. I think the regex only really addresses the specific case of delimiters between angle brackets though (which is fine if that's all you need! :) ).

Sure, we could patch the regex to try to find closing tags, but then what about * foo `bar* baz` boo*? If we keep special casing then we'd end up putting an entire copy of the parser in each inline.