Bug fix in FA, HWRF RRTMG and HWRF SAS

[climbfuji]

Update submodule pointer for fv3atm and date tag for regression test baseline in tests/rt.sh. See description in NCAR/ccpp-physics#447.

Associated PRs:

NCAR/ccpp-physics#447
NCAR/fv3atm#47
#45

For regression testing information, see below.

[climbfuji]

Regression testing on hera.intel against existing baseline using rt.conf: the following tests fail as expected in both REPRO and PROD mode:

fv3_ccpp_gfdlmp_hwrfsas
fv3_ccpp_regional_c768_FA
fv3_ccpp_regional_c768_HWRF_PBL
fv3_ccpp_regional_c768_FA_HWRF_RRTMG

rt_hera_intel_against_old_baseline.log
rt_hera_intel_against_old_baseline_fail_test.log

[climbfuji]

Creating new baseline and verifying against it (after moving to final location) on hera.intel: all tests pass

rt_hera_intel_create.log
rt_hera_intel_verify.log

[climbfuji]

Creating new baseline and verifying against it (after moving to final location) on cheyenne.gnu: all tests pass

rt_cheyenne_gnu_create.log
rt_cheyenne_gnu_verify.log

[climbfuji]

Creating new baseline and verifying against it (after moving to final location) on cheyenne.intel: "create" all tests pass, "verify" the following tests fail in both REPRO and PROD mode (known issue that is worked on) - ok for this commit.

fv3_ccpp_regional_c768_FA
fv3_ccpp_regional_c768_HWRF_PBL
fv3_ccpp_regional_c768_FA_HWRF_RRTMG

rt_cheyenne_intel_create.log

rt_cheyenne_intel_verify.log
rt_cheyenne_intel_verify_fail_test.log