`MeteoMapping.grid_mapping` logic by xugrid CentroidLocatorRegridder

This function:

https://github.com/Deltares/imod-python/blob/86bf0cab5431eae25e5413c9b5ae942e8887b0fa/imod/msw/meteo_mapping.py#L38

contains logic to map the SVAT to the coarse meteo grid files. This logic is the same as the one that can be found in the xugrid CentroidLocatorRegridder:

https://deltares.github.io/xugrid/api/xugrid.CentroidLocatorRegridder.html

The latter one has the advantage of working on unstructured grids as well, and should have more unit tests and the like as well. I recommend replacing the grid_mapping logic by roughly the following:

regridder = xu.CentroidLocatorRegridder(source=svat, target=meteo_grid)
weights = regridder.weights_as_dataframe()
index = weights["target_index"]

Note that index is an a 1D array for indexing along the faces. You can generate 2D arrays relatively easily, using numpy unravel_index: https://numpy.org/doc/stable/reference/generated/numpy.unravel_index.html

row, column = np.unravel_index(index, shape=svat.shape)

Then, repeat for the number of subunits, etc.

Start in imod/msw/meteo_mapping.py at the grid_mapping function and compare its current mapping logic with xugrid.CentroidLocatorRegridder. Review weights_as_dataframe(), target_index, and numpy.unravel_index for converting the one-dimensional face indices into row and column arrays. Done means the mapping uses the regridder approach, supports the stated subunit repetition, and preserves the expected mapping shape.