[bug] Repeatability issue for adupa_winds_220

[SamuelDegelia-NOAA]

Current behavior (describe the bug)

While testing PR #298, I noticed that the ctests sometimes (not always fails) due to a TestReferenceFloatMismatchError on the cost function printout for adpupa_winds_220. The Jo stats are changing for some runs which leads to the ctest failure. Here are a sample of five runs for fv3-jedi EnVar that only assimilate adupa_winds_220:

Run 1: 
  0: Test Line: 'CostJo   : Nonlinear Jo(adpupa_winds_220) = 3.2593295738966270e+04, nobs = 48714, Jo/n = 6.6907451120758443e-01, err = 1.6247685156989053e+18'
Run 2: 
  0: Test Line: 'CostJo   : Nonlinear Jo(adpupa_winds_220) = 3.2593295738966270e+04, nobs = 48714, Jo/n = 6.6907451120758443e-01, err = 1.6247685156989053e+18'
Run 3: 
  0: Test Line: 'CostJo   : Nonlinear Jo(adpupa_winds_220) = 3.2592904892550017e+04, nobs = 48714, Jo/n = 6.6906648792031076e-01, err = 2.2988790520072753e+18'
Run 5: 
  0: Test Line: 'CostJo   : Nonlinear Jo(adpupa_winds_220) = 3.2596883201982317e+04, nobs = 48714, Jo/n = 6.6914815457532362e-01, err = 2.7916051027288340e+00'
Run 4: 
  0: Test Line: 'CostJo   : Nonlinear Jo(adpupa_winds_220) = 3.2596328854534120e+04, nobs = 48714, Jo/n = 6.6913677494219570e-01, err = 2.2999878982948780e+18'

I noticed the crash for the rrfs_fv3jedi_2024052700_Ens3Dvar ctest so I am not sure if this is a problem specific to fv3-jedi or if it also occurs for mpas-jedi.

Expected behavior

The output CostJo stats should not change for each run.

Suggested Solution

Will start by investigating if this is related to individual filters.

[SamuelDegelia-NOAA]

It looks like this results from ObsErrorFactorPressureCheck and only occurs for fv3-jedi. I ran EnVar ten times for mpas-jedi and the cost function stats were the same every time. I am not sure why the problem occurs but for now I will just comment out this filter for the fv3-jedi ctests.