Update saber submodule

[SamuelDegelia-NOAA]

Now that Saber #928 has been merged, we can update that submodule in RDASApp. I will work on a PR for this. This will also (hopefully) resolve the ctest failures reported in #158.

[SamuelDegelia-NOAA]

The five mpas-jedi ctests are still failing with a reference mismatch error after updating Saber to the latest commit. It turns out that the reference files for mpas-jedi ctests are generated with CRTMv3 (RDASApp uses v2.4). This might explain why all the failing tests assimilate amsua data. There are some tests that pass with amsua data, though the differences might just be smaller than the float relative tolerance.

I am trying to confirm by updating to CRTMv3 and see if the ctests pass now, but the build is very slow due to needing to download the updated CRTM coefficients.

[guoqing-noaa]

Great find! @SamuelDegelia-NOAA

This poses another issue: do we want to upgrade to use CRTMv3? I think we should.

[SamuelDegelia-NOAA]

One potential issue with using CRTMv3 is that cross-validation against GSI might be harder. I am not sure if GSI is able to use CRTMv3 yet (or if it matters). But I'll tag @ShunLiu-NOAA, @xyzemc, @TingLei-NOAA if they want to provide other thoughts.

[xyzemc]

I don't think CRTMv3 is ready at this moment. As I know, global still keep using crtm2.4.0 for both GSI and JEDI.

[ShunLiu-NOAA]

We had discussion of the version of crtm. We have to stay with the version that used in GDASApp so that we can test satellite DA in RDASApp.

[guoqing-noaa]

@xyzemc Could you test the latest RDASApp to see whether the radiance DA works?
It looks like we may have different versions of CRTM for RDASApp and the jcsda mpasjedi tests.

[SamuelDegelia-NOAA]

Updating Saber and CRTM (to v3) allows four of the previously-failing ctests to pass. The GETKF test still fails but I will further investigate this tomorrow.

      Start 36: test_mpasjedi_3denvar_amsua_allsky
36/57 Test #36: test_mpasjedi_3denvar_amsua_allsky .............   Passed   11.62 sec
      Start 37: test_mpasjedi_3denvar_amsua_bc
37/57 Test #37: test_mpasjedi_3denvar_amsua_bc .................   Passed    9.35 sec
      Start 42: test_mpasjedi_4denvar_VarBC
42/57 Test #42: test_mpasjedi_4denvar_VarBC ....................   Passed   10.73 sec
      Start 43: test_mpasjedi_4denvar_VarBC_nonpar
43/57 Test #43: test_mpasjedi_4denvar_VarBC_nonpar .............   Passed   13.97 sec
      Start 52: test_mpasjedi_lgetkf_height_vloc
52/57 Test #52: test_mpasjedi_lgetkf_height_vloc ...............***Failed   16.78 sec
...
98% tests passed, 1 tests failed out of 57

Even though we plan to not update to CRTMv3 yet, it is good to know why those four tests were failing and that it is for an acceptable reason.