more general handling of precision

Anaconda on windows does not have np.float128 defined, but aliases
should work
https://numpy.org/devdocs/user/basics.types.html

python/damask/_grid.py

@@ -760,7 +760,7 @@ class Grid:
 
         """
         if fill is None: fill = np.nanmax(self.material) + 1
-        dtype = float if np.isnan(fill) or int(fill) != fill or self.material.dtype==np.float else int
+        dtype = float if isinstance(fill,float) or self.material.dtype in np.sctypes['float'] else int
 
         material = self.material
         # These rotations are always applied in the reference coordinate system, i.e. (z,x,z) not (z,x',z'')

python/damask/_vtk.py

@@ -246,8 +246,8 @@ class VTK:
                 raise ValueError('No label defined for numpy.ndarray')
 
             N_data = data.shape[0]
-            d = np_to_vtk((data.astype(np.float32) if data.dtype in [np.float64, np.float128]
-                      else data).reshape(N_data,-1),deep=True)                               # avoid large files
+            d = np_to_vtk((data.astype(np.single) if data.dtype in [np.double, np.longdouble] else
+                           data).reshape(N_data,-1),deep=True)                                      # avoid large files
             d.SetName(label)
 
             if   N_data == N_points:

python/setup.py

@@ -15,7 +15,7 @@ setuptools.setup(
     url='https://damask.mpie.de',
     packages=setuptools.find_packages(),
     include_package_data=True,
-    python_requires = '>=3.6'
+    python_requires = '>=3.6',
     install_requires = [
         'pandas>=0.24',                                                                             # requires numpy
         'scipy>=1.2',