Correct errors in metric calculations for DINO

[ma595]

This PR captures two errors in the calculation of the metrics:

• We had accidentally passed velocity_u into temperature_DWbox_metric, should have been temperature.
• The second error is more subtle:

Previously, we had renamed dims as so:

    if "x" in ds.dims:
        ds = ds.rename({"x": "nav_lon"})
    if "y" in ds.dims:
        ds = ds.rename({"y": "nav_lat"})

So this made the metrics (as they were given to us by Etienne) work fine.

However, when we select variables from the DataSet, for example,

temperature = ds.tn # giving a DataArray

The pre-existing dimensions (time_counter, nav_lev, nav_lat, nav_lon) are carried across as well as coordinates (time_counter, nav_lev). i.e. the nav_lat and nav_lon variables are not carried over. And note that the dims have been renamed to nav_lon and nav_lat.

To fix this, we need to promote the nav_lat(nav_lat, nav_lon) and nav_lon(nav_lat, nav_lon) to coordinates. The idiomatic way to do this is with the set_coords function:

    for name in ("nav_lat", "nav_lon"):
        if name in ds and name not in ds.coords:
            ds = ds.set_coords(name)  # zero-copy promotion to coordinate

Exploring xarray further, if Nemo had output nav_lat without the redundancy i.e. as 1D arrays. It would have been parameterised as so: nav_lat(nav_lat), which, on reading by xarray would automatically be recognised as a coordinate. The same is true of nav_lon.

Results

Running the following command:
python3 -m spinup_evaluation --sim-path ./tests/data/DINO_subsampled_A/ --mode restart

timestamp,check_density_from_file,check_density_computed,temperature_500m_30NS_metric,temperature_BWbox_metric,temperature_DWbox_metric,ACC_Drake_metric,ACC_Drake_metric_2,NASTG_BSF_max 
576000.0,0.000020,0.000032,3.236497,12.828081,0.007553,188.689106,188.689106,59.839730
576000.0,0.000020,0.000032,11.507896,5.197274,4.468086,188.689106,188.689106,35.515781

Metric	Old	New	Difference (new - old)
temperature_500m_30NS_metric	3.24	11.51	8.27
temperature_BWbox_metric	12.83	5.20	-7.63
temperature_DWbox_metric	0.0076	4.46	4.46
NASTG_BSF_max	59.84	35.52	-24.32

TODOs

• Parameterise relevant metrics with depth bounds
• Decide whether we need to rename x and y or enforce correspondence to indices from now on.
• Correct failing tests.
• There's also a mesh_mask path resolution fix which should probably be removed from this PR.

This closes #81

[ma595]

• run metrics on climatology of the true data.