Missed optimization in CellSetSingleType CellToPoint calc

In VTK, we optimize the CellToPoint construction for single-type cell arrays by simplifying the functors passed into ReverseConnectivityBuilder, but in VTK-m CellSetSingleType uses the same logic as CellSetExplicit, which is much slower as the offsets/sizes must be looked up instead of being implicitly defined.

In my benchmarks, this makes the algorithm take twice as long to run in VTK-m as it does in VTK.