[NodePartitionedMesh] Added two new data members for determining partition ID and the equation element ID
 for the Taylor-Hood element

Inverting some of the nested if-conditions. Now each vtkXxxArray type
is converted to some mesh property type (depending on the
architecture) or an error is raised.

Adding conversion of short = int16 types.

Merging conversion of vtkLongArray with vtkLongLongArray, and of
vtkUnsignedLongArray with vtkUnsignedLongLongArray.
For some systems the conversion of vtkLongArray, where the size of long
and long long types are different is not available. (But could be added
to the int type array conversion.)

Actually the use of vtkLongArray is discouraged by VTK: "The C++
standard does not define the exact size of the long type, so use of this
type directly is discouraged. If an array of 32 bit integers is needed,
prefer vtkTypeInt32Array to this class. If an array of 64 bit integers
is needed, prefer vtkTypeInt64Array to this class." -- from vtkLongArray
doxygen documentation.

Add compile-time tests for conversion of integer types since in OGS we
use the char, short, int, long types which are not guarantied to be of
corresponding length of vtk types.

The problem with the heap allocated array is that the ownership
is not clear. In the MeshEditing.Prism2Quad unit-test the
array of nodes was used after it has been already deleted
after element construction. Using array is safer and avoids
unnecessary heap alloctions.

Move implementation to cpp in anonymous namespace.

And with it all of the checks and recalculations are obsolete
and removed too.

Replace first part of the query with isBaseNode() as in
previous commits.
Use getNumberOfBaseNodes in calculation of "active" (regular) nodes.

The new implementation uses only local topology information to
distinguish base and "higher" order nodes.

The latter relies on local information of element types
and number of nodes, not the global numbering of nodes.