1. decrease branch number in CL code by replacing them into weights
2. decrease local mem pressure in reduce operation by using private variables
3. decrease image sampler pressure by caching data into local memory
4. remove unnecessary sync point on the HOST side.

Conflicts:
	modules/imgproc/src/opencl/precornerdetect.cl
	modules/nonfree/perf/perf_main.cpp
	modules/nonfree/perf/perf_surf.ocl.cpp
	modules/ocl/src/arithm.cpp
	modules/ocl/src/gftt.cpp
	modules/ocl/src/opencl/arithm_bitwise.cl
	modules/ocl/src/opencl/arithm_bitwise_binary_scalar_mask.cl
	modules/ocl/src/opencl/arithm_minMax.cl
	modules/ocl/src/opencl/arithm_nonzero.cl
	modules/ocl/src/opencl/arithm_sum.cl
	modules/ocl/test/test_arithm.cpp
	modules/viz/src/precomp.hpp
	modules/viz/src/vizcore.cpp
	modules/viz/src/vtk/vtkCloudMatSink.h
	modules/viz/src/vtk/vtkXYZReader.cpp
	modules/viz/src/vtk/vtkXYZReader.h
	modules/viz/test/test_precomp.hpp