Lake information into the GEOS BCS workflow

[biljanaorescanin]

This PR introduces HydroLAKES-TopoCat v1.1 (2023) lake and reach information into the GEOS BCS workflow.
This update does not alter tile geometry or Pfafstetter indexing. It augments tile metadata with a single encoded LakeTopoCat/ReachTopoCat touch type field.

It implements:

1. Lake and reach preprocessing pipeline added under preproc/lake/v1 with Python scripts and README.
2. Tile-level LakeTopoCat/ReachTopoCat touch type appended to tile .nc4 files through updates to rmTinyCatchParaMod and related makebcs calls.

@gmao-rreichle

[gmao-rreichle]

@biljanaorescanin : I tried to clean up a bit. See also comments/questions below.

[gmao-rreichle]

Do we really need REAL64 here? By construction, there are only 10 distinct real numbers in the 30 arcsec files: [0, 0.11111, 0.22222, ..., 0.88888, 1]. Even for a tile that spans a 1-deg by 1-deg area, there are still only ~10,000 raster (30 arcsec) grid cells. Single-precision should be able to accommodate the digits required to handle the useful information.

[biljanaorescanin]

Done in this commit: d0753af

[mathomp4]

@gmao-rreichle In re last commit, what constants aren't defined in use netcdf? They should all be there, though maybe as their NF90_ variants.

If some are missing, I can talk to the devs.

[gmao-rreichle]

@mathomp4 : For the record, I'm attaching the build log (0-2.txt) that I got when I replaced "include 'netcdf.inc'" with "use netcdf". Looks like nothing with "NF_" is available from "use netcdf".
I tried this based on ChatGSFC saying that "use netcdf" should be backward compatible and should make the F77 API available. I guess ChatGSFC was hallucinating...
I suppose we could change the NF_* to the NF90_* equivalent, but that's more than I'm willing to do at this time. I just used the F77 API again...

[biljanaorescanin]

The last commit addresses all @gmao-rreichle comments. I reran the workflow after the updates. As expected, changing tile_lake_frac from REAL64 to REAL32 introduces small round-off differences relative to the previous outputs, so the files are not zero-diff.
The roundoff in M09 nc4 tile file:
Date Time Level Gridsize Miss Diff : S Z Max_Absdiff Max_Reldiff : Parameter name
15 : 0000-00-00 00:00:00 0 2163601 0 440376 : F T 2.9802e-08 5.9464e-08 : tile_lake_frac

[biljanaorescanin]

With last commit I've implemented the LakeTopoCat/ReachTopoCat option as one encoded output variable, tile_lake_type, with values -9999, 0, 1, 2, 3 for UNDEF, no touch, lake-only, reach-only, and lake+reach. It is defined for typ==0/19/20 and set to -9999 for typ==100.

The implementation reads lake_presence_any and reach_presence_any from the 30″ H/V preprocessing outputs and maps them to tile space via tile_id.
I validated the regular EASE M09 output against the standalone Python prototype; LakeType counts and typ==19 keep/remove counts match.