Adds stand alone test_MOM_EOS and time_MOM_EOS

[adcroft]

• Added simple single-thread program to invoke EOS unit tests
• Added not-as-simple program to time the cost of calculate_density() and calculate_spec_vol() for both scalar and array APIs
  • Placed in new directory config_src/drivers/timing_tests/
• Renamed MOM_unit_test_driver.F90 to test_MOM_file_parser.F90
• Updated .testing/Makefile
  • Added list of programs in config_src/drivers/unit_tests - These are added to BUILDS if DO_UNIT_TESTS is not blank. (DO_UNIT_TESTS was an existing macro but it might be unneeded) - These programs are compiled with code coverage
  • Added list of programs in config_src/drivers/timing_tests
    • These programs are compiled with optimization and no coverage
  • Fixed rule for building UNIT_EXECS (which did not re-build properly because the central Makefile was trying to model the dependencies even though those dependencies are in the build/unit/Makefile.dep)
  • Added convenient targets build.unit, run.unit, build.timing, run.timing
• Timing tests currently time a loop over 1000 calls (so that the resolution of the CPU timer is not too large) and 400 samples to collect statistics on timings. On gaea c5 this takes about 10 seconds.
  • The results are written to stdout in json.
• Added placeholder build and run of timing_tests to GH workflow.
  • Enabled for [push,pull_request]
  • We probably will not be able to use timings from GH but I still want to exercise the code we know the timing programs aren't broken by a commit.
• Also added driver for string_functions_unit_tests

[codecov]

Codecov Report

Merging #516 (b6a5976) into dev/gfdl (0f2a69d) will decrease coverage by 0.01%.
The diff coverage is 28.88%.

❗ Current head b6a5976 differs from pull request most recent head aa22fd8. Consider uploading reports for the commit aa22fd8 to get more accurate results

@@             Coverage Diff              @@
##           dev/gfdl     #516      +/-   ##
============================================
- Coverage     37.45%   37.44%   -0.01%     
============================================
  Files           270      270              
  Lines         79692    79722      +30     
  Branches      14823    14824       +1     
============================================
+ Hits          29849    29855       +6     
- Misses        44308    44331      +23     
- Partials       5535     5536       +1     

Files	Coverage Δ
...ig_src/drivers/unit_tests/test_MOM_file_parser.F90	92.00% <100.00%> (ø)
...ameterizations/lateral/MOM_mixed_layer_restrat.F90	70.03% <ø> (ø)
src/equation_of_state/MOM_EOS.F90	54.36% <0.00%> (-0.89%)	⬇️
src/framework/MOM_error_handler.F90	56.86% <40.74%> (-2.91%)	⬇️

📣 Codecov offers a browser extension for seamless coverage viewing on GitHub. Try it in Chrome or Firefox today!

[Hallberg-NOAA]

I have examined the Fortran90 portions of this commit, and they all make sense to me. I would approve the changes to the 8 Fortran90 files in this commit (after one minor change to some indents), but I would like to have someone else (@marshallward?) review the new or revised Makefile and .py and .yml files in this PR.

[marshallward]

Python script needs a PEP8 update, but otherwise looks fine.

I noted some Fortran style fixes.

There are some conflicts from the recent Makefile changes.

[Hallberg-NOAA]

This is all looking good to me now, and it also looks like Marshall's concerns have been addressed.

[Hallberg-NOAA]

This has passed pipeline testing at gitlab.gfdl.noaa.gov/ogrp/MOM6/-/pipelines/21302.