Switch from mpich to openmpi in GitHub workflows

[adcroft]

GitHub switched images from focal to jammy a few days ago. Since then linking to libFMS.a has been failing. I'm not sure why, but using OpenMPI instead of mpich works. I suggest making this switch to revive the GH workflows.

[codecov]

Codecov Report

Merging #270 (ea694a2) into dev/gfdl (d01c42a) will increase coverage by 0.03%.
The diff coverage is n/a.

@@             Coverage Diff              @@
##           dev/gfdl     #270      +/-   ##
============================================
+ Coverage     37.09%   37.13%   +0.03%     
============================================
  Files           263      263              
  Lines         73395    73434      +39     
  Branches      13675    13674       -1     
============================================
+ Hits          27225    27268      +43     
- Misses        41137    41142       +5     
+ Partials       5033     5024       -9     

Impacted Files	Coverage Δ
src/tracer/MOM_generic_tracer.F90	0.00% <0.00%> (ø)
src/parameterizations/CVmix/cvmix_kpp.F90	0.00% <0.00%> (ø)
src/ice_shelf/MOM_ice_shelf_diag_mediator.F90	0.00% <0.00%> (ø)
src/core/MOM_barotropic.F90	58.98% <0.00%> (+0.01%)	⬆️
...parameterizations/vertical/MOM_diabatic_driver.F90	46.37% <0.00%> (+0.07%)	⬆️
src/framework/MOM_restart.F90	32.76% <0.00%> (+0.12%)	⬆️
src/core/MOM_dynamics_split_RK2.F90	62.94% <0.00%> (+0.15%)	⬆️
src/framework/MOM_diag_mediator.F90	58.83% <0.00%> (+0.18%)	⬆️
src/framework/MOM_checksums.F90	58.04% <0.00%> (+0.47%)	⬆️
src/framework/MOM_io.F90	33.26% <0.00%> (+0.49%)	⬆️
... and 3 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[marshallward]

As an OpenMPI fan, I support this :).

Though we should probably test both MPICH and OpenMPI once we figure out the underlying problem.