Change from "html.HTMLParser" to "etree.XMLParser" for validation

[jedie]

html.HTMLParser doesn't accept "unknown" html tags.
e.g.: It raised an error because of <nav> tag :(

The etree.XMLParser seems to be a better choice: It raise errors on really broken html documents,
but accepts all tags.

Another difference: HTMLParser accepts whitespaces in closing tags like: </ \r\n h1> and
XMLParser not.

Thsi PR also doesn't activate recover=False, because with this option, validate django admin
default templates doesn't work. But they seems that they are fine html codes...

Another thing:

etree.XMLSyntaxError will raise a very, very big traceback and not any context to the broken HTML
document.
This is a problem in combination with snapshot tests: Because it's totally unknown what part of the
HTML document contains the error.

Now we get really helpful messages, that points to the error and contains some context lines, e.g.:

StartTag: invalid element name, line 5, column 25
--------------------------------------------------------------------------------
02     <foo>
03         <bar>
04             <h1>Test</h1>
05             <p> >broken< </p>
----------------------------^
06             <p>the end</p>
07         <bar>
08     </foo>
--------------------------------------------------------------------------------