Cross-references from proofs to theorems

[Rob Beezer]

I've implemented cross-references from "detached" proofs to the theorem they
claim to prove. (PROOF-LIKE to THEOREM-LIKE, really.)

Semantics are the same as for an "xref", just attributes on the "proof" element
(@ref to @xml:id, @text to control the style of the text).

A bit rough right now, but nobody will exercise this code unless they add the
@ref onto a proof. Extensive testing suggests no harm. Please experiment.

Note that we have three types of "proof":

1. Inside hint, answer, solution. A "solution proof". Not numbered, not
stylable, @ref not effective. Limited.

2. Inside a THEOREM-LIKE. An "inner proof" (thanks, Alex). The usual.

3. Everywhere else. A block that is just a proof. "Detached" (or standalone).
Needs a number, I think - I have no good idea what is happening there, I'll
catch it with the big numbering refactor.


See Theorem 21.3 in the sample article, and a detached proof almiost immediately
afterwards, but for an intervening paragraph.

LaTeX - look at this first. No cheating. Placement seems OK to me.

HTML - Placed inside another knowl link, a bit of a mess. And *also* placed in
the body/content of the knowl. (not final, experimenting!) I'm at a bit of a
loss here. How should this go/look? Note there are two cases for the heading:
"Proof" is a knowl, or not. Plus the "proof" might have a title, which will
replace "Proof".

Note: I'm avoiding the use of the word "of", just to avoid having it in every
localization, but maybe that should just happen.

I'll clean this up, and the new PROOF-LIKE, plus update the schema, before
announcing broadly.

Rob

[kcri...@gmail.com]

Thanks for this - I think it will be a highly valued feature in the long run!  Though current authors may take longer to have time to use it :)