elf: Add test case for [BZ #19329]

Test concurrent dlopen and pthread_create when the loaded modules have
TLS.  This triggers dl-tls assertion failures more reliably than the
nptl/tst-stack4 test.

The dlopened module has 100 DT_NEEDED dependencies with TLS, they were
reused from an existing TLS test. The number of created threads during
dlopen depends on filesystem speed and hardware, but at most 3 threads
are alive at a time to limit resource usage.

---
v5:
- moved from nptl/ to elf/
- use many tls modules from another test (affects the timing a bit as
  these need more relocation processing, but still detects the race)
- use xdlopen
- use stdatomic.h
- moved from tests-internal back to tests
- use acq/rel atomics instead of relaxed, this works too and cleaner.
- moved after bug fix patch.

v4:
- rebased, updated copyright year.
- moved to tests-internal because of <atomic.h>

v3:
- use the new test support code.
- better Makefile usage so modules are cleaned properly.

v2:
- undef NDEBUG.
- join nop threads so at most 3 threads exist at a time.
- remove stack size setting (resource usage is no longer a concern).
- stop creating threads after dlopen observably finished.
- print the number of threads created for debugging.