Python 3 bugfix in ccpp-framework, ESMF 8.1.0bs27, cleanup rt_utils.sh, fix RRTMGP regression tests

[climbfuji]

Description

This PR:

• contains a submodule pointer update for fv3atm for a Python 3 bugfix in ccpp-framework
• uses the correct Python interpreter to run ccpp_prebuild.py from CMakeLists.txt
• removes machine dependent logic from tests/rt_utils.sh
• updates ESMF from 8.1.0bs21 to 8.1.0bs27 on all tier-1 platforms, and all tier-2 systems that I am responsible for
• fixes the RRTGMP regression tests (these tests were running the original RRTMG schemes due to typos in the regression test files); however, since these currently crash (@dustinswales is working on it), they are temporarily turned off in tests/rt.conf and tests/rt_gnu.conf
• minor updates of pull request template
• turn on INLINE_POST for hera.gnu, cheyenne.intel and cheyenne.gnu in gnumake build (they are already on in the cmake build)

The update from ESMF 8.1.0bs21 to ESMF 8.1.0bs27 (unexpectedly) changes the results of the regression tests. No changes in the actual results, but in the metadata of the phyf*.{nc,nemsio} and dynf*.{nc,nemsio} files:

• order of global and variable attributes
• the time variable (counting hourse since forecast initialization) sometimes - but not always, seems to depend on the regridding option - changes from double to float

Testing

• full regression tests passed for new baseline (20200914) on hera.intel, hera.gnu, cheyenne.intel, cheyenne.gnu, orion.intel, wcoss_cray, wcoss_dell_p3
• code compiles successfully on jet.intel when compile_cmake.sh is used; however, when using rt.sh, there is a problem with line 215 in FV3/ccpp/physics/CMakeLists.txt; investigation ongoing, will be fixed in a forthcoming commit
• code compiles and runs on macOS with clang+gfortran and gcc+gfortran
  STILL TODO/CONFIRM
• manual tests will be run on gaea.intel, stampede.intel
• tests should also be run on the new wcoss platform (which shares the PBS scheduler logic with cheyenne in tests/rt_utils.sh)

Dependencies

If testing this branch requires non-default branches in other repositories, list them.
Those branches should have matching names (ideally)

Do PRs in upstream repositories need to be merged first?
If so add the "waiting for other repos" label and list the upstream PRs

• waiting on Python 3 bugfix in ccpp-framework (update submodule pointer only) NOAA-EMC/ufsatm#171
• waiting on scripts/ccpp_prebuild.py: bugfix for Python 3 NCAR/ccpp-framework#322

[climbfuji]

Note: temporarily added tests/rt_rrtmgp.conf (by mistake), will remove before the PR is merged.

[MinsukJi-NOAA]

Utests can be run on Hera, Orion, Dell with Intel compilers (I had some trouble with ecFlow on Cray)

[climbfuji]

Should I add this information to the PR template?

[MinsukJi-NOAA]

Since regression tests are supported on all platforms (tier 1, 2, etc.), maybe it would be helpful to add that utests are only supported on tier 1 (Cray will have to come later)? e.g. unit tests (utests on tier 1 platform)

[climbfuji]

@MinsukJi-NOAA thanks for the docker/ci updates, the automated tests all passed

[climbfuji]

@junwang-noaa @DusanJovic-NOAA this PR is ready from my side. Waiting for CI tests to complete.