Combine all tests into a single suite to increase parallelism. NFC

[sbc100]

My motivation for this change is to allow for increased parallelism that is not limited to how
many tests are in a give module. I'm hoping to do more splitting up of test_other.py, for
example, and I don't want to loose parallism.

After this change all tests are combined into a single parallel test suite. For example:

$ ./test/runner jslib other.test_a* core0.test_a* --skip-slow
Test suites: ['test_core', 'test_jslib', 'test_other']
Running test_core: (88 tests)
Using 88 parallel test processes
[88/88] test_asan_vector (test_core.core0.test_asan_vector) ... ok

----------------------------------------------------------------------
Ran 88 tests in 3.296s

OK (skipped=5)
Running test_jslib: (54 tests)
Using 54 parallel test processes
[54/54] test_jslib_aliases_closure_wasm64 (test_jslib.jslib.test_jslib_aliases_closure_wasm64) ... ok

----------------------------------------------------------------------
Ran 54 tests in 5.287s

OK
Running test_other: (41 tests)
Using 41 parallel test processes
[41/41] test_abspaths (test_other.other.test_abspaths) ... ok

----------------------------------------------------------------------
Ran 41 tests in 5.210s

OK
Total core time: 228.881s. Wallclock time: 13.795s. Parallelization: 16.59x.
====================

TEST SUMMARY
    test_core: 88 run, 0 errors, 0 failures, 5 skipped
    test_jslib: 54 run, 0 errors, 0 failures, 0 skipped
    test_other: 41 run, 0 errors, 0 failures, 0 skipped

After:

$ ./test/runner jslib other.test_a* core0.test_a* --skip-slow
Running 183 tests
Using 128 parallel test processes
[183/183] test_abspaths (test_other.other.test_abspaths) ... ok

----------------------------------------------------------------------
Ran 183 tests in 7.459s

OK (skipped=5)
Total core time: 301.490s. Wallclock time: 7.459s. Parallelization: 40.42x.

Note the wall clock time is less since we ran all the tests in parallel.

If you try to mix parallel and non-parallel test module you now get an error. e.g.:

$ ./test/runner core0 benchmark
runner: error: attempt to mix parallel and non-parallel test modules

[kripken]

how do these two variables differ? from the name and from the code I'm having a hard time figuring it out.

[sbc100]

The first module we find dictates if we are parallel or not.

If we later find a module that does not agree, we error out. I will update the names to be more clear.