cmems_nc_to_ini produces netcdf file incoherent depth variable

[veenstrajelmer]

It contains positive-up values (so negative is below water), but a positive-down attribute. This is the way to work around https://issuetracker.deltares.nl/browse/UNST-7455. However, what also seems to work is to provide positive down with positive values, like in an old file like this: p:\11210334-004-dcsm-fm\models\dflowfm3d-noordzee_0_5nm-j22_6-v1a\initial_conditions\CMEMS\CMEMS_itdate19951225_20041225b.nc.

Todo:

• test with positive down attr and positive down values, so coherent file again. >> this also gives faulty results.
• test with raw cmems data, could be that renaming affects standard name. depth is always positive down, so maybe we should have z instead. Also consider if the standard_name has to be updated when renaming the variable in ds_apply_attrs()
• prevent incoherent depth variable
• prevent nan values in file (bffill/ffill is on depth but was renamed to z)

Update 15-8-2024: the comparison in UNST-7455 went wrong because the layer definition was incorrect. It should have been with layertype=2 instead of 1. That way, the comparison makes more sense. When using positive down attr with positive up values, all layers get the value from the top layer, this is incorrect. When using positive up, or using the raw cmems data more directly, the values over depth correspond to the expected values.

Todo:

• use the more direct approach to make this more transparent, alternatively just remove the overwriting of the depth attribute.
• improve the tests to make sure the above is covered
• update whatsnew