Invalid attribute bug in DocumentFragment parsing

Marc and I had a conversation last year about this bug (http://sourceforge.net/tracker/index.php?func=detail&aid=1995218&group_id=195122&atid=952178). Basically, if you try to parse the following text in a DocumentFragment parser, you'll get an error:

<a - href="/">link</a>
<a . href="/">link</a>

The error is as follows:

> Caused by: org.w3c.dom.DOMException: INVALID_CHARACTER_ERR: An invalid or illegal XML character is specified.

This is because the attributes the XML specification is very strict about special characters in attribute names. Running this through the regular DOMParser doesn't because the parsing logic is different.

Anyway, I have attached a patch that creates a new feature, "http://cyberneko.org/html/features/enforce-strict-attribute-names", that does not process XML-illegal attribute names for DocumentFragments. It is off by default at Marc's request. I hope that everyone benefits from this patch.

I have also attached a non-JUnit test case. You will have to make sure the path of the file name in the DOMParser test points to the HTML file that I also attached to make it work in your environment. That particular test is not very important because all it does is show that the DOMParser does not have the same flaw.

Basically, this is the same bug as reported before, but I've introduced a test case and a patch.