use the same approach like in fisheye calibration: instead of setting
masked out rows to zero, remove them from the equation system.
This way JtJ does not become singular and we can use the much faster LU
decomposition instead of SVD.

This results in a speedup of the Calibrate unit tests of 3x-10x.

WinRT parallels library is causing conflicts in VC++ due to inconsistent usage with C++/CX which is now present in parallels.cpp.  This support should be included in all projects with WinRT specific code except if doing special native C++ tasks.

Fix C++/CX support for all WINRT and bring back native-mode only support which was removed

so it matches the code.

related issues: #5423 #5476

PR #2968: cce2d992 8578f9c5
Fixed bug which caused crash of GPU version of feature matcher in stitcher

The bug caused crash of GPU version of feature matcher in stitcher when
we use ORB features.

PR #3236: 5947519f
Check sure that we're not already below required leaf false alarm rate before continuing to get negative samples.

PR #3190
fix blobdetector

PR #3562 (part): 82bd82e2
TBB updated to 4.3u2. Fix for aarch64 support.

PR #3604 (part): 091c7a38
OpenGL interop sample reworked not ot use cvconfig.h

PR #3792: afdf319f
Add -L for CUDA libs path to pkg-config

Add all dirs from CUDA_LIBS_PATH as -L linker options to
OPENCV_LINKER_LIBS. These will end up in opencv.pc.

PR #3893: 122b9f8e
Turn ocv_convert_to_lib_name into a function

PR #5490: ec5244a7
fixed memory leak in findHomography tests

PR #5491: 0d5b739d
delete video readers

PR #5574

PR #5202