...when we know we have a working compiler. This will reduce
the risk of faulty edits by exposing them rather than hiding
them as "unresolved". It also harmonizes behavior with that of
run_sim_test.
* c/c.exp: Mark C tests failing compilation test errors.