Strange diff for SCREAMv0 DP test

[rljacob]

After PR #6113 was merged, the SMS_Ln9_P24x1.ne4_ne4.FDPSCREAM-ARM97 test showed a few diffs from baselines in the eam.h0 file.

SUMMARY of cprnc:
 A total number of    243 fields were compared
          of which     11 had non-zero differences

The variables were QDIFF, TDIFF, and *_RES variables.

The PR gave the case 2 new namelist variables and values:
pgrad_correction = 1
hv_ref_profiles = 6

Logically, this change should be BFB for this case.

@bogensch noted:

that is odd as all those variables (QDIFF, TDIFF, *RES) are computed from the state. So it's weird that one set would be b4b and the other not.... One commonality of the output variables listed that are not b4b is that they are all computed in the components/eam/src/dynamics/se/se_iop_intr_mod.F90 file. This provides a clue but looking at the PR (and the namelist changes ....), I can't explain off of the top of my head why these diffs are present. I will have to chew on this.