Regrid out of core

See merge request !82

Merge branch 'regrid-out-of-core' of https://gitlab.com/deltares/imod/imod-python into regrid-out-of-core

iMOD-wq has some annoying inconsistenties when it comes to setting concentrations. It accurately checks the number of active elements when concentration is given by IDF, but not when given by scalar.

In case of a scalar, it allocates a FULL array, the entire domain. This means that crch_t1_p?_l? allocates nlayer * nrow * ncolumn elements. To avoid this, the methods figure out which layers require concentrations, and only writes those to the runfile.

MXSS is also computed accurately for scalar concentrations, by multiplying the number of layers by the full number of row and columns.

Fixes #119.

* It's absolutely essential to call .sel() BEFORE putting it into dask.delayed
* Slicing of source should occur on both all available dimensions
* Fix slice_source negative starting index
* Make regridder._regrid pure
* Special case chunks that are 0-sized along one or more dimensions; just fill with fill_value
* Introduce extra overlap attribute, for example required by linear interpolation, at it reaches over cell boundaries.

Add chunking to tests; parametrize chunksize; identical -> allclose(values) because of cellsize coordinates

* Update voxelize to work with simpler _coord function
* Create _increasing_dims function
* Update tests, get rid of decreasing coords in low level function tests

lifecycle note, badges and Deltares logo

See merge request !81