Fix bug in chgres_cube subroutine search_many

[LarissaReames-NOAA]

DESCRIPTION OF CHANGES:

There was a bug with an unallocated array being used on all PETs in a call to search_many when it only needed to be used on the head node. I fixed this is in all calls to search_many subroutine, including the unit test for this routine.

TESTS CONDUCTED:

Consistency and unit tests, with code compiled in debug mode, passed on Jet.

DEPENDENCIES:

N/A

DOCUMENTATION:

No documentation changes needed.

ISSUE:

Fixes #797

CONTRIBUTORS (optional):

N/A

[LarissaReames-NOAA]

All of the tests appear to be failing in places that have nothing to do with anything I've changed, so I'm not really sure how to rectify that. This certainly goes back to our discussions in our last management meeting about having so many tests that constantly need updating but that aren't adding much value.

It looks like all of the GitHub actions still use nceplibs which we no longer use for modules, instead using hpc-stack, so some library versions required by UFS_UTILS aren't available. Should we create an issue to update the actions to source and build hpc-stack instead?

[GeorgeGayno-NOAA]

All of the tests appear to be failing in places that have nothing to do with anything I've changed, so I'm not really sure how to rectify that. This certainly goes back to our discussions in our last management meeting about having so many tests that constantly need updating but that aren't adding much value.

It looks like all of the GitHub actions still use nceplibs which we no longer use for modules, instead using hpc-stack, so some library versions required by UFS_UTILS aren't available. Should we create an issue to update the actions to source and build hpc-stack instead?

Don't worry about the failing tests. They are failing for reasons other than errors in our codes. @arunchawla-NOAA is aware of the problem.

[GeorgeGayno-NOAA]

Does argument field_data_2d need to be an argument of search_many? The routine does a FieldGather and FieldScatter on that array. Can it be declared locally? Because the routine expects field_data_2d to be dimensioned as i_target and j_target. And on most tasks, that array is dimensioned as (0,0). I wonder if that would cause problems.

[GeorgeGayno-NOAA]

Checked out the branch at 36fc338 on WCOSS2. The doxygen built with no warnings. All regression tests passed.

Then, recompiled the branch using 'Debug' mode. All regression tests completed without error. There were very small differences with the baseline data, which were likely due to the different compiler options used in 'Debug' mode. Importantly, the fix to search_many is working.