FV3 cold-start experiments crash with SIGSEGV

[nikizadehgfdl]

ATM cold-start experiments crash with SIGSEGV as follows:

    forrtl: severe (174): SIGSEGV, segmentation fault occurred
    Image              PC                Routine            Line        Source
    fms_am4p0_compile  0000000000A27633  test_cases_mod_mp        1599  test_cases.F90
    fms_am4p0_compile  00000000007F91B5  fv_restart_mod_mp         324  fv_restart.F90
    fms_am4p0_compile  00000000004ABD9F  atmosphere_mod_mp         288  atmosphere.F90
    fms_am4p0_compile  000000000047FB4C  atmos_model_mod_m         802  atmos_model.F90
    fms_am4p0_compile  000000000040A370  coupler_main_IP_c        1713  coupler_main.F90
    fms_am4p0_compile  0000000000402485  MAIN__                    612  coupler_main.F90

I had the following comments from Lucas a while back:
"The fourth dimension of q is not ncnst but instead ncnst - pnats where pnats is the number of diagnostic tracers. Unfortunately thi
s assumption is made in many cases in the code.
A quick fix would be to change the call to test_cases() in fv_restart.F90, replacing 'ncnst' with 'ntprog'.
However a long-term fix is needed throughout the core to ensure that the
distinction between advected, non-advected, and diagnostic tracers is
made clear and avoids future memory problems.
"