Tweaked test/bench id globbing to avoid duplicating cases

Before, globs that match both the suite name and case name would cause
end up running the case twice. Which is a bit of a problem, since all
cases contain their suite name as a prefix...

  test_f* => run test_files
             |-> run test_files_hello
             |-> run test_files_trunc
             ...
             run test_files_hello
             run test_files_trunc
             ...

Now we only run matching test cases if no suites were found.

This has the side-effect of making the universal glob, "*", equivalent
to no test ids, which is nice:

  $ ./scripts/test.py -j -b '*'  # equivalent
  $ ./scripts/test.py -j -b      #

This is useful for running a specific problematic test first before
running the all of the tests:

  $ ./scripts/test.py -j -b test_files_trunc '*'

scripts/bench.py

@@ -816,6 +816,7 @@ def find_ids(runner, bench_ids=[], **args):
             bench_ids__.extend(suite
                 for suite in expected_suite_perms.keys()
                 if fnmatch.fnmatch(suite, name))
+            if not bench_ids__:
                 bench_ids__.extend(case_
                     for case_ in expected_case_perms.keys()
                     if fnmatch.fnmatch(case_, name))

scripts/test.py

@@ -833,6 +833,7 @@ def find_ids(runner, test_ids=[], **args):
             test_ids__.extend(suite
                 for suite in expected_suite_perms.keys()
                 if fnmatch.fnmatch(suite, name))
+            if not test_ids__:
                 test_ids__.extend(case_
                     for case_ in expected_case_perms.keys()
                     if fnmatch.fnmatch(case_, name))