EARL patch

Validator EARL shows traces of an RDF learning curve: XML striping
is suppressed for no obvious reason, and overuse of
rdf:parsetype="Resource" generates huge numbers of bnodes.
Looking at the report generated by a page with a small number of errors
(http://valet.webthing.com/invalid.html) I can't make too much sense of
the picture generated by the RDF validator from the HTML validator report.

I attach a patch to generate a compact but meaningful EARL report.
I'm not quite sure how it'll go down with the EARL folks, but IMO
it's worth putting to them.

Rationale:
  (1) Validator asserts that {page} passes or fails {valid HTML}
      That is one assertion - no problem.
  (2) For the individual error messages, I've minimised the Subjects
      and Objects, on the grounds that they are known to be {points
      within the page} and {violations of the HTML standard} respectively.
      Not quite sure whether everyone will be happy with the minimisation,
      but it looks nice to me!

-- 
Nick Kew