Based on a specialized definitions parser for the HTML spec built to improve on the existing scraping done in shepherd.

This particular patch doesn't cover all the identified dfn-type that could be set, only the ones that were easy to treat (i.e. where the data-x and the dfn occur on the same line).

The setting of dfn-type and dfn-for is based on the WebIDL defined in the spec

This change:

* replaces any data-dfn-type=element, etc., attributes with ones literally named
  "element", etc.

* replaces any data-dfn-for attribute with one literally named "for"

* replaces data-export="" and data-noexport="" attribues with value-less
  attributes literally named "export" and "noexport"

* drops "export" from any dfn that has a dfn type attribute