Add lake tile type; update Landice ("glc") collection to match M21C; update CI, turn off spack build

[biljanaorescanin]

This PR is part of 3 PRs that help us run just lake tiles in GEOSldas offline mode.

1. Add Lake offline mode.
2. Add lake HISTORY ( 1D and 2D collections) and Restart. (NOTE: maybe 1D collection should be removed if we don't enable EASE grid runs.)

Additionally, updates Landice HISTORY output to match M21C plus key ISSM outputs.

Also updates CI, turns off spack build.

Regression and functionality testing:

• 0-diff tested for run tile_type 100 20 (zero diff to current global and globalcs tests).
• Tested for tile_type 100 19 20. Test was done for c360 only.

PR should also work for Lakes in EASE grid tile space, but GEOS_Util still needs remap_restart functionality for Lake restarts in EASE grid tile space. Without this remap_restart functionality, we don't have an EASE Lake restart (unless it's assembled manually, which is too painful for testing).

Related PRs:

• make ISSM GridComp name consistent with LANDICE etc GridComp names for NUM_LDAS_ENSEMBLE>1 GEOSgcm_GridComp#1510 (merged)
• Enable offline use of Lake GC; add analytical derivative of exchange coeffs for Lake and Landice GEOSgcm_GridComp#1470 (merged)
• Include Lake GridComp in LDAS sparse checkout; update CI and turn off spack build GEOSldas#862 (merged)

[gmao-rreichle]

@biljanaorescanin, @weiyuan-jiang, @lcandre2, @agstub :

I did my best to fix up some missing functionality in the GEOSldas GC portion of the Lake PR. Specifically, we need Lake to run with the Land ensemble.

We already decided to run only one member for Landice. Lake doesn't have the same computational constraints as Landice+ISSM, but I think we should also limit Lake to one member. After all, if we added an ensemble of Lake GCs, it would not be used at all (no assimilation or uncertainty output). So I deleted the hard stop triggered by "with_lake==.true. .and. N_ens>1" and tried to make Lake behave like Landice as far as ens members are concerned. @weiyuan-jiang, please review the code changes.

It may not have been technically needed, but I added a LakeAvg GC. @lcandre2, I think the [Land,Lake,Landice]Avg GridComp give us an opportunity to change LONG_NAME attributes from those in the GCM GC, and I took advantage of that for Lake.

I forget if we figured out how the "native" ISSM outputs (i.e., the "issm" collection) will behave when we run in ensemble mode. It's probably best to just try it out. I also didn't try to use the "HAVE_ISSM" preprocessor directive in the LDAS GC. @agstub, I may need some help if this doesn't work as is.

@biljanaorescanin : Please try to set up a (c90?) test run for a few days with NUM_ENSEMBLE=4 (and no data assimilation for now). Write out the "lnd", "lake", "glc", and "issm" collections defined in GEOSldas_HIST.rc. The "source" GridComp specs might need to change to LAKEAVG, LANDICEAVG, etc to make sure we're pulling from the "AVG" Grid Comps and not the LAKE, LANDICE, etc GCs. (This will need to be cleaned up in the script that preprocesses GEOSldas_HIST.rc) Also, for now you can just make four identical copies of a single restart and use them to restart the ensemble (or it may still be possible to restart an ens simulation from a single member restart that is then used for all members). I'm sure I'm forgetting lots of details, so please be in touch with questions. Thanks!

Also, see my comments below.

[weiyuan-jiang]

quote: "So I deleted the hard stop triggered by "with_lake==.true. .and. N_ens>1" and tried to make Lake behave like Landice as far as ens members are concerned."

That looks fine with me.

[biljanaorescanin]

All requested regression tests passed.

[gmao-rreichle]

@mathomp4 : Now that #194 is in, I restored ifx in the workflow (GEOS-ESM/GEOSgcm_GridComp#1510). I hope I got this right. When you get a chance, please look over the CI changes one more time to see if they're how would want them to look before we merge the PR (which we're ready to do once Scott has merged GCM GC PR 1510). Thanks