Thread local as dynamic so that threads will communicate status

[GavinMendelGleason]

The overall run_tests should fail if any test fails, but that requires that we coordinate book-keeping through the various predicates passed/5, failed/4, failed_assertion/7 etc. For now I've just re-declared them as dynamic rather than thread_local.

[JanWielemaker]

Hmmm. There was a reason for the thread_local declarations. I was wondering what that could be. I think I know. One is to deal with test runs initiated by (re-)loading files which may happen concurrently. The other is for calls such as below to stress-test test suites that are supposed to be thread-safe.

?- concurrent_forall(between(1, 4,_), run_tests)).

Seems this needs to be a bit more complicated, such that the helpers know the thread they are supporting and recording the (main) test thread-id with the passed, etc facts.