* Correct IDs in the learning map domain to use type NMTOKEN instead of
type ID
* Classification domain: add processing-role to the 3 topicref
specializations, default=resource-only (this was added to DITA 1.2 after
the domain was created, and the update didn't make it to this domain)
* subjectScheme: set the default for processing-role to "resource-only" on
the root element
* Fix the desing pattern for constraints so that it's more modular.
NOTE - This is new and uses files entities to pull in the content into