#1713 Added the possibility of setting user_block_size manually for gpu::matchTemplate function (and gpu::convolve). Added a buffer param into these functions. Removed using of 2^n block sizes when it's not necessary.

fixed gpu module error reporting
added asynchronous version of some functions

fixed bug in gpu::matchTemplate (added normalization routine to make the GPU version consistent with the CPU one), added test cases from the ticket #1341

fixed BFM warning ("cannot tell what pointer points to")

added assertion after all kernels calls

updated normalization routine in the matchTemplate to avoid division by zero on black images (ticket #798), added test

moved crossCorr (as NPP_Staging wrapper) into public GPU module part from the internal matchTemplate files

added support of multichannel images into gpu::matchTemplate (all methods except CCOEFF based), refactored

added support of CV_TM_CCORR (via FFT) into gpu::matchTemplate (versions both with block and without blocks)

module reorganization: added folder with pure device functions, cuda_shared.hpp renamed to internal_shared.hpp

added first version of gpu::matchTemplate, currently it works only with 8UC1 images and SQDIFF method