* __shfl_up_sync with proper mask value for CUDA >= 9

* BlockScanInclusive for CUDA >= 9

* compatible_shfl_up for use in integral.hpp

* Use CLAHE in cudev

* Add tests for BlockScan

YUV/YCrCb conversions rewritten to wide intrinsics (#13649)

* YUV: minors

* YUV42x conversions template-merged

* more template-merged YUV42x conversions; some NEON code removed

* rgb2yuv<float> vectorized

* yuv2rgb<float> vectorized

* memcpy removed

* Yuv2RGB<ushort> vectorized

* unused code removed

* rgb2yuv<ushort> vectorized

* rgb2yuv<uchar> vectorized

* v_pack_u used (up to +30% perf)

* yuv2rgb<uchar> vectorized

* fixed compilation

This reverts commit 3721c8bb
except changes in modules/dnn/test/test_tf_importer.cpp

* Add Operator override for multi-channel Mat with literal constant.

* simple test

* Operator overloading channel constraint for primitive types

* fix some test for #13586

- Before this PR, following tests failed on some platform.
  CUDA_OptFlow/FarnebackOpticalFlow.Accuracy/19
  CUDA_OptFlow/FarnebackOpticalFlow.Accuracy/23

- The algorithm now recognizes the OPTFLOW_USE_INITIAL_FLOW flag.
  Previously, when the flag was set, it did not use the flow data
  passed as input, instead used some garbage data in memory.

- More strict test limit.

* fixed tutorial code of py_matcher

* fixed imread mode

* Fix the bug in case determinant of rotation matrix is -1

* calib3d(test): check det(R) == 1