Reproducibility problem with cnmatrix first occurs in dev172 going from ccs_config_cesm0.0.88 to ...89

A cnmatrix reproducibility problem first occurs in dev172 and more specifically when going from ccs_config_cesm0.0.88 to ccs_config_cesm0.0.89. This is the only difference that I see when comparing 89 to 88:

`git diff ccs_config_cesm0.0.88 ccs_config_cesm0.0.89`

```diff
diff --git a/machines/derecho/config_machines.xml b/machines/derecho/config_machines.xml
index 633fe74..6f9d6fd 100644
--- a/machines/derecho/config_machines.xml
+++ b/machines/derecho/config_machines.xml
@@ -21,11 +21,11 @@
get_local_rank
TRUE

- mpiexec
+ mpibind

--label
- --line-buffer
- -n {{ total_tasks }}
+ --line-buffer
+ --



```

_Originally posted by @slevis-lmwg in https://github.com/ESCOMP/CTSM/issues/640#issuecomment-2176742021_

Definition of Done:
- [x] Add warning to build-namelist to abort when matrix is on and OMP_NUM_THREADS > 1 and test it
- [x] Have our tests setup to ignore the warnings for our tests for this configuration and test it
- [ ] Add a test to the build_namelist_unit_tester with OMP_NUM_THREADS = 2 and verify that it aborts as intended
- [ ] Use NTHRDS_LND rather than OMP_NUM_THREADS
- [ ] Fix CN Matrix code so that OMP_NUM_THREADS env variable is NOT used, but use values from the NUOPC cap
- [ ] Make sure that fixes it -- and if not track down the problem
- [ ] Fix the real problem
- [ ] Remove the warning in the namelist