Read and broadcast data from MPI root rank during init and timestep init phase in GFS time vary

[grantfirl]

SOURCE: @grantfirl

DESCRIPTION OF CHANGES:

• Changes to SCM to work with Read and broadcast data from MPI root rank during init and timestep init phase in GFS time vary; add time vary interstitials for NEPTUNE ccpp-physics#1187
• Also needs [scm/dev] Update scm/dev from main (includes ccpp_bcast PR + followup NOAA fixes) ccpp-physics#1209

ISSUE: None

ASSOCIATED PRs:
NCAR/ccpp-physics#1209

TESTS CONDUCTED: SCM RTs

REGRESSION TEST CHANGES: None expected

[grantfirl]

@scrasmussen @hertneky @dustinswales There is an unexpected difference in one gfortran/release/dev test. The plots show that the differences start near machine precision, indicative of an optimization-related result difference. Plus, the debug runs show bit-for-bit. I think it's OK to merge.

[dustinswales]

@scrasmussen @hertneky @dustinswales There is an unexpected difference in one gfortran/release/dev test. The plots show that the differences start near machine precision, indicative of an optimization-related result difference. Plus, the debug runs show bit-for-bit. I think it's OK to merge.

@grantfirl If the DEBUG are b4b, then I completely agree. Also, the Intel tests are b4b.
Also, it's a v15 ntiedtke test, so not a huge impact.