WUI based implementation to initUndistortRectifyMap (#14994)

* Add initUndistortRectifyMap performance test

* Move cv namespace boundaries

* Add wide universal intrinsics based implementation to initUndistortRectifyMap

* Dispatch undistort

* Fix dumpToFile

* Add test

* Fix test

avoid generation of copy ctors with unique_ptr members

G-API: clean up accuracy tests (#14945)

* Delete createOutputMatrices flag

Update the way compile args function is created

Fix instantiation suffix print function

* Update comment (NB)

* Make printable comparison functions

* Use defines instead of objects for compile args

* Remove custom printers, use operator<< overload

* Remove SAME_TYPE and use -1 instead

* Delete createOutputMatrices flag in new tests

* Fix GetParam() printed values

* Update Resize tests: use CompareF object

* Address code review feedback

* Add default cases for operator<< overloads

* change throw to GAPI_Assert

G-API: Add output allocation tests for backends (#15012)

* Add output tests for backends

* Fix large size test: output is in fact reallocated

* Use cv::Mat copies for reallocation tracking

* Separate LargeSizeWithCorrectSubmatrix test

* Rename backed output allocation tests

* Address code review feedback

Update test names

Add illustrative "expect (non-)empty" checks

Rename mat "copy" to mat reference

Add more pointer checks

* Add illustrative checks

* Fix G-API export specifier in standalone build

* Make dummy GAPI_EXPORTS in case of non-OpenCV builds

* Add old version under #if 0

* Reduce store gathering pressures - speeds thresholds by up to 20%

* Rename temporary histogram array and initialize so that MACOSX builder is happy

 - Added new graph compile time argument to specify multiple independent
ROIs (Tiles)
 - Added new "executable" with serial loop other user specified
ROIs(Tiles)
 - refactored graph traversal code into separate function to be called
once
 - added saturate cast to Fluid AddCsimple test kernel

- writeRaw(): support structs
- readRaw(): 'len' is buffer limit in bytes (documentation is fixed)

- fixes crashes in face-detector application
- libc++_shared.so is bundled automatically via CMake helper project

* Support Deconvolution3D on IE backend

* Add test tag

* Fix tests

may cause crashes on ARM platform

* Support Conv3D on OCV backend

* Add header

* Add perf tests

* Support pool3d

* Enable Resnet34_kinetics on OCV backend

* Add test

* Fix conv

* Optimize Conv2D