Possible memory leak in elm/crops

When #8494 was first merged to next, the MEMLEAK TOLERANCE on Chrysalis was set to 5% and it reported a leak in 3 of the long running crop model tests:

SMS_Ly2_P1x1.1x1_smallvilleIA.IELMCNCROP.chrysalis_intel.elm-fan
memleak detected, memory went from 728.360000 to 770.520000

SMS_Ly5_P1x1.1x1_smallvilleIA.IELMCNCROP.chrysalis_intel.elm-force_netcdf_pio
memleak detected, memory went from 600.370000 to 642.000000

SMS_Ly5_P1x1.1x1_smallvilleIA.IELMCNCROP.chrysalis_intel.elm-per_crop
memleak detected, memory went from 609.480000 to 641.680000

Two long running test did NOT report a memory leak. One with and one without crops:
SMS_Ly1.ELM_USRDAT.I1850CNPRDCTCBC.chrysalis_intel.elm-kilocraft (only 1 year)
SMS_Ly2_P1x1.1x1_smallvilleIA.I20TRGSWCNPCROP.chrysalis_intel.elm-lulcc_sville (different config from other 3)

These are barely over 5%. We later raised the tolerance to 10% and the fail went away but the underlying leak is still there.