choosing different methods for different variables #16
Description
At the moment, most libraries (xesmf, xarray-regrid, xarray's interpolation methods) use one method for all variables in a dataset – though xesmf falls back to nearest for categorical values, I believe.
However, that's not actually what we'd like to do: different kinds of data need different interpolation methods, and although it is possible to do this manually by splitting into smaller datasets and recombining after the regridding, this is the kind of work I'd expect a library to do for me.
In order to do that, I believe we need to come up with some kind of way to make specifying the methods per variable as simple as possible (also, we need more methods to actually make sense, but that's a different question).