Bugfix GF convection, Thompson cleanup 2019/07/09

[climbfuji]

This PR and associated PRs listed below fix a bug in the GF convection scheme introduced in a previous code cleanup PR (#251) and cleans up/improves code for the CCPP Thompson MP scheme.

• physics/GFS_PBL_generic.F90: add missing tracers for Thompson MP to tracer advection array for PBL schemes
• physics/cu_gf_deep.F90: bugfix, replace exit statements in do loop with cycle statements
• physics/cu_gf_driver.F90: cleanup, instead of passing in convective_transportable_tracers array, only pass in the components required
• physics/cu_ntiedtke.F90, physics/samfdeepcnv.f, physics/samfshalcnv.f: add notes to check dimensions of clw array
• physics/module_mp_thompson.F90, physics/mp_thompson.F90: cleanup, return with an error if CCN_ACTIVATE.BIN is not found, remove temporary stopping of model

Testing: logs will be posted below

Associated PRs:
https://github.com/NCAR/NEMSfv3gfs/pull/199
https://github.com/NCAR/FV3/pull/176
#283

[climbfuji]

Testing on MacOSX/GNU: bit-for-bit identical results through restarts for both the GSD_v0 suite and the GSD_noah (i.e. GSD suite with Noah LSM instead of RUC LSM).

[climbfuji]

NEMSCompsetRun test on Theia against official baseline: all pass (including one CCPP test, but no GSD suite test).

NEMSCompsetRun.log

[climbfuji]

CCPP-GSD tests (create and verify, including restart runs) all PASS on Theia/Intel

rt_ccpp_gsd_verify.log
rt_ccpp_gsd_create.log

[climbfuji]

CCPP-GMTB tests (create in static mode, verify in static and dynamic mode) all PASS on Theia/Intel

rt_ccpp_gmtb_verify.log
rt_ccpp_gmtb_create.log

[climbfuji]

Regression tests against official baseline on Theia/Intel PASS except fv3_ccpp_gsd tests - this is to be expected because of the changes for Thompson MP and the bugfixes for GF convection.

rt.log

This PR is ready for review and merging.

[grantfirl]

Looks good to me. Again, sorry for the loop exit bug!