Add a test that verifies that when ctest -j 4 is called, at most, 4 tests
are executed at any one time.
The test works by running the same script as each of 100 tests. And then setting
up test properties for DEPENDS, RUN_SERIAL, PROCESSORS and COST in order to get
the tests to run in a semi-deterministic ordering, even in parallel. The script
writes a file, sleeps for a bit, and then deletes the file. In the meantime, it
counts files that currently exist, and emits output that triggers a test failure
if the count of files is ever greater than 4.
Prior to the commit that fixed bug #12904, this would result in a failed test
because the output of some of the tests would indicate that more than 4 tests
were running simultaneously.
Now that this issue is resolved, this test will help guarantee that it stays
resolved moving forward.