":load M" now complains if M.hs doesn't contain module M, but ':load
M.hs' does not.  This seems reasonable to me, but there's potential
for confusion.  I'm not sure how to improve the situation though.

- RankN1 was a bogus test.  There error didn't have anything to do with
  rank-n types, but was just due to an ambiguous signature.

This makes things more robust when running with threads

Its really a testsuite-framework bug that these mentions of 'opt' have
lingered so long.  They should have been rejected as a non-existent way.