add cardsort; premise has order attr

[wrigjl]

Add cardsort as an option for a matching problem. Also, allow premise to have an order attribute.

(Be gentle with me please, this is the first time I've meddled in the affairs of XML validation).

[rbeezer]

First, apologies for taking so long to get to this. We now have cardsort (old matches) and the slightly more general matching.

I edits your commit some for cardsort, and it still has your name on it (along with an AI co-author). Then I added another commit with similar schema stuff for matching.

Be gentle with me please

Always! Next comment will be from AI, mostly for the discussion about the script that automates making the derived files. I suspect that is in a README somewhere as well.

[rbeezer]

Two commits merged to address this PR and extend the schema.

Commit 1 (amended from your original): Adds <cardsort> to the Matches production, removes the deprecated <matches> element and @order on <match>, allows multiple <premise> per <match>, and switches the content model for premise and response in cardsort from mixed {BlockText?} to TextLong so that inline elements like <m> validate.

Commit 2: Adds <matching> as an alternative to <cardsort> in the Matches production, with <premise> (optional @ref) and <response> (@xml:id, optional @order), also using TextLong content.

About derived files. The schema source of truth is schema/pretext.xml, which is a literate program. The .rnc and .rng files are derived from it by running xsltproc with xsl/pretext-litprog.xsl to extract the compact RELAX-NG (.rnc), then trang to convert to RELAX-NG XML (.rng). The script schema/build.sh automates this. PRs that modify the schema should edit only pretext.xml and include the regenerated derived files.