To set image geometry (origin, spacing, directions) developers had to
call 3 methods. This was:
- inconvenient: most imaging software stores image geometry
  in a 4x4 homogeneous matrix and need to disassemble the information
  to use the 3 methods
- inefficient: Modified event is invoked 3 times, so if there are
  observers that react to data change they are triggered 3x for a
  single change
- may lead to temporary inconsistencies: while setting the geometry,
  Modified event is called twice while the image data has still
  incorrect geometry, neither the previous nor the new one,
  but a mix of the two.

To resolve this, two new methods are added to vtkImageData:

    void ApplyIndexToPhysicalMatrix(vtkMatrix4x4* source);
    void ApplyPhysicalToIndexMatrix(vtkMatrix4x4* source);

"Apply" is used instead of "Set" in the method name
to make it clear that the source matrix is not stored
as object in the image data but its content is used for
updating the origin, spacing, and direction matrix.