Reader memory usage/memory usage when wrapping vtkDataSets

As pointed out in https://github.com/pyvista/pyvista-support/issues/500#issuecomment-921108703 and https://github.com/pyvista/pyvista-support/issues/500#issuecomment-921106561, readers seem to duplicate memory usage.

Using this code shows that the mesh data is copied from the reader to the mesh object. This is only released when deleting the reader object.

import pyvista as pv
from pyvista import examples

from memory_profiler import profile

@profile
def run():
    filename = examples.download_parched_canal_4k(load=False)
    reader = pv.get_reader(filename)
    mesh = reader.read()
    del reader

if __name__ == "__main__":
    run()

$python test.py 
Filename: test.py

Line #    Mem usage    Increment  Occurrences   Line Contents
=============================================================
     6    110.2 MiB    110.2 MiB           1   @profile
     7                                         def run():
     8    110.2 MiB      0.0 MiB           1       filename = examples.download_parched_canal_4k(load=False)
     9    110.4 MiB      0.2 MiB           1       reader = pv.get_reader(filename)
    10    302.8 MiB    192.3 MiB           1       mesh = reader.read()
    11    206.8 MiB    -96.0 MiB           1       del reader

[MatthewFlamm]

It turns out that some datasets, like PolyData, use shallow copy by default with an option. Others only allow deep copy. This seems like a straightforward PR at this point.

[MatthewFlamm]

It fixes the issue raised in this PR, which is that wrapping certain vtk datasets doubles memory and may seem to leak memory unless there vtk dataset is garbage collected. This manifests in the reader since the reader keeps a reference to the vtk dataset object.