Release consistent with GEOSadas-5_27_1_p3 CVS Tag as of 2021-Feb-25

This release is equivalent to the GEOSadas-5_27_1_p3 CVS tag as of 2021-Feb-25.

It is also equivalent to the git tag cvs/GEOSadas-5_27_1_p3

Fix for GFDL Microphysics

@aoloso found a bug in FV_StateMod.F90 that (luckily) is in a path not normally taken. This release fixes this bug, and is zero-diff for "normal" runs of GEOS.

Fix for MAPL module use

A very minor bugfix to CubeGridPrototype.F90 where it was doing:

    use MAPL_RegridderManagerMod, only: regridder_manager
    use MAPL_RegridderSpecMod, only: REGRID_METHOD_BILINEAR	

and with modern MAPL we should just use MAPL for all access.

Update for fv3.j script

This release updates the fv3.j script to extend the LD_LIBRARY_PATH much like as is done in gcm_run.j to use install/lib to find shared-object libraries built by GEOS.

Update CI and remove Externals.cfg

This is a patch release with the following changes:

1. Update CI image and improve CI workflow
2. Remove Externals.cfg with move to use mepo

Bug fix for ntracers > 11

As discovered by @lizziel of Harvard, there is an issue in FV3 if the number of tracers is greater than 11 (and 11 only works because of luck). To wit, the code was casting an integer into a string with i1.1:

        write(myTracer, "('Q',i1.1)") ntracer-1	

For n < 11, this is fine and even n=11 this is okay since it becomes Q*. But, after that, everything is Q* and...not sure what happens.

So, the format is extended to i5.5 which should be safe for a long time.

Updates to allow ioserver with Standalone

This is a zero-diff change to FVdycoreCubed_GridComp that only affects the fv3.j and fv3_setup scripts.

The changes allow FV3 Standalone to actually use the IOServer. Before the support was essentially broken.

Also adds CircleCI testing.

Update to fv3_setup

This is a minor patch to use cp instead of rsync in fv3_setup as some CI images might not have rsync. As all we are doing is copying a single file, cp works just as well and is universal.

Lower domains_stack_size in fv3.j

As found with testing FV3 standalone on non-NASA machine, the default value of domains_stack_size in fv3.j was unnecessarily large. This lowers it to match the default in GEOS GCM.

Bug fixes for fv3.j script

This patch release fixes a few bugs in the fv3.j.

Nothing in FV3 core itself was touched, so it is trivially zero-diff for that. As for running fv3.j it is tested zero-diff as well. The changes are mainly for better MPI support and to fix a couple bugs.