TOC should remove all html entities from the parsed headers before moving on, to handle WYSIWYG-generated html

Hi!

Just encountered this use-case with a customer:
She had made headers correctly, but with hr-tags just above them, which TinyMCE made into

<h2><hr />Blabla</h2><code/> and which made TOC choke.
I've fixed it in her case (by manually moving all the hr /'s), but I think it would be good if TOC automagically removed all non-text elements from between the <h*>-tags before parsing further, to make sure that no problems arise.

(The only tags that might be useful to keep are strong> and em, which can add emphasis to the TOC. Otherwise I think all tags should be removed. I would guess that this is already partly done, so that for instance the usecase <code><h2><a href="google.com">Foo</a></h2>

does not result in a double link, but unfortunately have not had the time to test.