[AC21.5] Fix typo in test of equalp, add more tests

[Stephen J. Turnbull]

Aidan Kehoe writes:

 > It would be perfectly reasonable to do it, too. I spent a chunk of yesterday
 > procrastinating and wrote the below, and was going to write tests for *it*
 > before I committed it, but more tests are unlikely to go astray.

Looks reasonable, but note that not everything here is a pure
optimization, some are only heuristic wins since for some cases they
involve additional calls at runtime for the test.  I wonder if the
ones involving characters are worth it.  (For future reference, since
you've already written and nothing looks like a pessimization.)

 > Also, note that the string equalp tests (both in my patch and in the actual
 > function) use the general case table, while the character equvalence tests
 > use the case table of the current buffer when they call #'downcase.

This stuff has always bothered me but I don't have an answer.