Allow increment files with time dimension to be recentered

[DavidNew-NOAA]

This PR allows the increment variables with a time dimension to be read in by recentersigp.x.

Currently the FV3-JEDI "auxgrid" IO interface, used in the Global Workflow when running JEDI, writes the ensemble member increments with a singleton time dimension. recentersigp.x will not read these increments in, because it checks for variables with 3 dimensions, rather than 4. This PR allows either 3 or 4 dimension increment variable to be read.

[DavidNew-NOAA]

@RussTreadon-NOAA I re-ran enkfgdas_ecen with this fix, and now the ensemble member increments and their recentered counterparts are no longer identical.

Also, I don't have permission to add reviewers to this PR.

[RussTreadon-NOAA]

Good catch! Changes look to me. We do not have gsi_utils ctests so I'll test the proposed changes in GSI and JEDI-based g-w CI.

Approve pending successful g-w CI results.

[RussTreadon-NOAA]

WCOSS2 (Cactus) cycled test

Conduct the following test on Cactus

1. Install g-w develop at 380946c. This installation used gsi_utils.fd @ f716012
2. Run g-w CI. All jobs run to completion. Save output. This is the CONTROL
3. Replace sorc/gsi_utils.fd with bugfix/ndims_recenter at 85b129f.
4. Run g-w CI. All jobs run to completion. Save output. This is the UPDATE

GSI-based DA
Comparison of UPDATE and CONTROL 20211221 06Z enkfgdas output for C96C48_hybatmDA (GSI-based) shows that the two sets of recentered enkfgdas analysis increment files are identical. This is expected. This PR should not alter recentersigp.x output from GSI-based parallels. Additionally, the unrecentered (atminc) and recentered (ratminc) analysis increment files differ. This, too, is expected. recentersigp.x should alter the increment file.

JEDI-based DA
Comparison of UPDATE and CONTROL 20240224 00Z enkfgdas output for C96C48_us_hybatmDA (JEDI-based) shows that the two sets of recentered enkfgdas analysis increment files differ. This is expected. The CONTROL recentered analsis increment files are not correct. The CONTROL unrecentered and recentered analysis increment files are identical. This is wrong. The recentered increment file should differ from the unrecentered file. This is true in the UPDATE. The UPDATE unrecentered and recentered analysis increment files differ. This is correct.

The recentersigp.x created by this PR properly processes GSI and JEDI based enkfgdas analysis increment files.

[DavidNew-NOAA]

Thanks @RussTreadon-NOAA