Harvest option landuse_timeseries run fails on perlmutter for intel compiler in debug mode

[glemieux]

The e3sm.log file shows the following:

 65: forrtl: severe (194): Run-Time Check Failure. The variable 'edloggingmortalitymod_mp_loggingmortality_frac_$CUR_HARVEST_TAG' is being used in '/global/u1/g/glemieux/E3SM-project/e3sm/components/elm/src/external_models/fates/biogeochem/EDLoggingMortalityMod.F90(344,33)' without being defined
 65: Image              PC                Routine            Line        Source
 65: e3sm.exe           0000000003F37C73  edpatchdynamicsmo         265  EDPatchDynamicsMod.F90
 65: e3sm.exe           000000000376DCBF  edmainmod_mp_ed_e         217  EDMainMod.F90
 65: e3sm.exe           0000000000B3929D  elmfatesinterface        1252  elmfates_interfaceMod.F90
 65: e3sm.exe           00000000009E44F6  elm_driver_mp_elm        1296  elm_driver.F90
 65: e3sm.exe           00000000009631AB  lnd_comp_mct_mp_l         617  lnd_comp_mct.F90
 65: e3sm.exe           00000000004B0923  component_mod_mp_         757  component_mod.F90
 65: e3sm.exe           000000000046BEEA  cime_comp_mod_mp_        2963  cime_comp_mod.F90
 65: e3sm.exe           00000000004991CB  MAIN__                    153  cime_driver.F90
 65: e3sm.exe           0000000000435CDD  Unknown               Unknown  Unknown
 65: libc-2.31.so       000015459AA3C24D  __libc_start_main     Unknown  Unknown
 65: e3sm.exe           0000000000435C0A  Unknown               Unknown  Unknown

Per the lnd.log, this appears to be happening in conjunction with the logging event.

Note that this does not occur for the exact same test on derecho (intel compiler in debug mode) nor does it occur on perlmutter with the debug mode gnu test, so we have some coverage overlap of this mode at least.

I think the issue is that harvest_tag needs to be set to zero in the LoggingMortality_frac here:

fates/biogeochem/EDLoggingMortalityMod.F90

Lines 286 to 290 in 825579d

	if (hlm_use_lu_harvest == ifalse) then
	! 0=use fates logging parameters directly when logging_time == .true.
	! this means harvest the whole cohort area
	harvest_rate = 1._r8

Looking back on #1116 I realized this had come up and I even verified this fix via bb3d58e, but for some reason I reverted the fix with a later commit.