Grid mask support

added 4 commits

b8b12795 - Grid: fix 1D/2D structured vtk export
1966af67 - Grid: fix 1D/2D structured vtk export for rectilinear
3e54b7e1 - Grid: correctly reshape mask when casting grid to unstructured
0921babc - deps: update pyevtk version with crucial bugfix

added 1 commit

980bda8e - Grid: add methods to convert compressed data

Add nodata_value / missing_value to grid class and use a default of -9999 (integer to be compatible with all numeric types).

mentioned in issue #110

Some decisions form Martin, Tom and me:

always use NaN to indicate masked values in data arrays
we can't use masked numpy arrays, since they have a deep issue with pint: https://github.com/hgrecco/pint/issues/1776
problem is integer arrays, since NaN only works with float, other libraries have same problem (xarray, pandas)
we need to add dtype information to the Info container which should be float by default #109

added data feature labels

added 1 commit

c747b4ea - grid: add data_points_compressed method

Compare with previous version

added 2 commits

88686118 - Grid: respect pint when in from_compressed; better mask setting
0ec9aeff - IO: add 'compressed' flag to push_/pull_data

Compare with previous version

@mlange this is the implementation I was talking about. Basic idea is, that model developers could use the flag compressed within their model-components when pulling data from Inputs and pushing data to Outputs. All other components don't change their behavior.

Intended usage:

    def _update(self):
        self._time += self._step

        a = self.inputs["A"].pull_data(self.time, compressed=True)
        b = self.inputs["B"].pull_data(self.time, compressed=True)

        result = a * b

        self.outputs["Prod"].push_data(result, self.time, compressed=True)