pysteps.utils with xarray

[dnerini]

Include new xarray-based data model into the pysteps.utils module (See #12).

utils.conversion

• remove metadata arguments (since all attributes are now part of the input xr.DataArray object)
• make all methods available as xarray accessors for DataArrays (e.g. can use both the data_array.pysteps.to_rainrate() or pysteps.utils.to_rainrate(data_array) syntax)
• to_reflectivity() has the option to convert to dB [dBZ] or keep linear scale [Z]
• add type hints for inputs and outputs
• reduce code duplication
• adapt tests

utils.dimension (DEPRECATED)

• deprecate aggregate_fields_time in favor of xarray's coarsen method.
• deprecate aggregate_fields_space in favor of xarray's coarsen method.
• deprecate aggregate_fields in favor of xarray's coarsen method.
• deprecate clip_domain in favor of xarray's sel method.
• deprecate square_domain since it is obsolete.

utils.images

• morph_opening takes xr.DataArray as input, and it also handles cases with time dimension "t"
• argument n of morph_opening is renamed kernel and can either be the size of the kernel or the kernel itself.
• update tests

utils.interpolate

• all interpolation methods take and return xarray's object (both Dataset and DataArray)
• migrate deprecated scipy's Rbf to new RBFInterpolator (scipy>=1.7)
• update tests

utils.transformation

• remove metadata arguments (since all attributes are now part of the input xr.DataArray object)
• make all methods available as xarray accessors for DataArrays (e.g. can use both the data_array.pysteps.sqrt_transform() or pysteps.utils.sqrt_transform(data_array) syntax)
• add general back_transform method to remove any existing data transformation.
• boxcox_transform and db_transform use a simpler offset to avoid computing log(0), instead of having to define the threshold and zerovalue parameters as in V1 (closes dB_transform ignores existing metadata #198)
• refactor nq_transform: in particular, the inverse transformation needs an empirical target distribution (the template argument).
• add type hints for inputs and outputs
• reduce code duplication
• rewrite tests

[dnerini]

OK let's say this is ready for review!

• still failing tests (need to adapt all methods that use the utils module)
• didn't touch at the utils.spectral module for now, do we want to adopt xarray for complex data too? this would be similar to what is done in xrft. Proposition: should we adopt xrft ? @pulkkins

I know it is a lot of changes, I don't expect you to go through all of them. Rather, please try to see whether you agree with the general changes in the interfaces listed in the summary at the top.

Thanks!

[RubenImhoff]

This is an amazing amount of work, @dnerini! I'll try to have a look at it soon.

[RubenImhoff]

Amazing work, @dnerini! Because you've done so much work, I can't promise that I've spotted all possible mistakes in the code, so let's also wait for another review by one of the others. ;)

[RubenImhoff]

Very minor: I believe we have:
# -*- coding: utf-8 -*-
in most of the scripts