Address #2426 — do not allow elements or classes inside a datatype

[sydb]

What I did:

• Added para to "TD-datatypes" (22.5.3.1) asserting that, while they may be used in a content model, because datatypes are for attr vals, cannot have element or attr.
• Changed example that shows defining “more than one” occurrence of a datatype from @target of att.pointing to @columns of <layout>. I did this because I think the URI example is particularly problematic, as someone who does not carefully read a modern version of teidata.pointer but knows how xs:anyURI is defined could be quite confused.
• Made clear in list of ways a <dataRef> refers to a dataype that TEI datatypes are defined by the <dataSpec> element.
• Added a role="warn" constraint to <dataSpec>.
• Corrected half a dozen references to XML Schema Part 2: Datatypes Second Edition in dataRef and dataFacet tagdocs.

Note that I used an @role of "warn" on the constraint itself, on the theory that we should provide a sort of deprecation period. Feel free to disagree. (Either way — no deprecation needed, remove "warn", or re-word warning to make it a real deprecation.)

[sydb]

The ticket suggests we add prose that says, essentially “Macros are for use in content models, not attr values”. I did not do so, because <macroRef> cannot occur inside <attDef>, anyway.

[raffazizzi]

Reviewed files and tested schematron constraint. It all looks good to me. Build checks should be fixed before merging.

[sydb]

The build check failures are “rnv: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.34' not found (required by rnv)”, so I kinda doubt it has anything to do with this PR. How do we ask Mr. Jenkins to re-run the checks? (Do I just go in and change some whitespace somewhere? :-)

[HelenaSabel]

Everything looks great. Build failed because of #2455. Tests pass in my local machine so I think it’s safe to merge.