(cherry picked from commit ff002203)

use new abstract interface and hidden implementation

use new abstract interface and hidden implementation

use new FastFeatureDetector abstract interface and hidden implementation

used BufferPool mechanism instead

* remove overloads with explicit buffer, now BufferPool is used
* added async versions for all reduce functions

As opencv_xfeatures2d is part of opencv-contrib and not opencv repository,
${OpenCV_SOURCE_DIR}/modules/modules/include is not a correct include
path - use ${opencv_xfeatures2d_SOURCE_DIR}/include instead

The FGD background subtraction model was part of opencv_legacy module;
as it is gone now, it makes little sense to benchmark the GPU version,
either, so the whole test is removed.

The header is gone now, along with the opencv_legacy module.

Env setup for testing package implemented using /etc/profile.d;
Variable with path for all native samples added;
Path for test binaries and test data updated.
(cherry picked from commit 39201e68)

(cherry picked from commit f332cba1)

Env setup for testing package implemented using /etc/profile.d;
Variable with path for all native samples added;
Path for test binaries and test data updated.

(cherry picked from commit b75cbfde)

Conflicts:

	cmake/OpenCVModule.cmake

Allow to build samples via OpenCV binaries from distribution package
(find_package with OpenCV_DIR).

Conflicts:
	modules/core/include/opencv2/core/cuda.hpp
	modules/cudacodec/src/thread.cpp
	modules/cudacodec/src/thread.hpp
	modules/superres/perf/perf_superres.cpp
	modules/superres/src/btv_l1_cuda.cpp
	modules/superres/src/optical_flow.cpp
	modules/videostab/src/global_motion.cpp
	modules/videostab/src/inpainting.cpp
	samples/cpp/stitching_detailed.cpp
	samples/cpp/videostab.cpp
	samples/gpu/stereo_multi.cpp