A series of bug fixes including: 1) gradient calculation error;
2) addressing boundary edges interpolation in certain rare cases;
3) setting scalar values (if requested) in one fill_n operation;
4) correct treatment of normals.

Also the documentation was improved, and a new .cxx test case was
added.

Volumes with large empty spaces (i.e., areas with no iso-contour) can
b skipped over. This skipping process was not always working and
has been fixed.

Bill Lorensen authored Mar 06, 2015 and Will Schroeder committed Sep 11, 2015

Some Mac compilers fail to compile member function calls with the
redundant prefix:
vtkFlyingEdges2DAlgorithm<T>:: and vtkFlyingEdges3DAlgorithm<T>::.

Removing (the redundant) prefix resolves the compilation issues and
seems to be portable.

Also replaced static const unsigned char with enums. Some Macs had
trouble resolving the static consts at link time.

All templated code has been consolidated into a single templated
class. The vtkSMPTools::For() has been rewritten to loop over slices
rather than edges (a tad faster and the code is cleaner).

Trying different looping strategies.

Each of the three passes of the algorithm has been integrated
into the vtkSMPTools. This enables TBB and X-Kaapi support (as
well as well as sequential and simple vtkMultiThreader
capabilities). Note that there is no distinction between a serial
and parallel version of FlyingEdges (i.e., separate implementations).