New computeECC function, and updated findTransformECC function to make gaussian filtering optional (#13837)

* fix for https://github.com/opencv/opencv/issues/12432 with doc and tests

* Added doc string for new parameter.

* Fixes suggested by Alalek for getting around ABI incompatibility.

* Update to docstring, to remove parameter that isn't relevant.

* More updates based on Alalek's usggestions.

Resize reworked using wide universal intrinsics (#13781)

* Added wide universal intrinsics optimized implementation for 3 channel bit-exact linear resize

* Reworked linear resize using new wide LUT intrinsics

* Fix for VSX intrinsics

All <arm_neon.h> includes in core/cv_cpu_dispatch.h are protected by an
ifndef __CUDACC__ to prevent attempting to use neon intrinsics when
compiling cuda kernels (.cu) -- this prevents hard errors such as
  error: identifier "__builtin_neon_qi" is undefined

Add this same protection to flann/dist.h to fix compilation involving
flann.hpp.

Fix arguments parsing. Add possibility to choose between different features type. Add keypoints matching visualization. Auto format code.

- gemm: keep baseline only (lapack is 10x+ faster, lets reduce binary size)
- transform / distTransform
- scaleAdd (32f/64f only)
- Mahalanobis: keep baseline only (no perf tests)
- mulTransposed: keep baseline only (no perf tests)
- dot

- Fix hanging issue on 2080 Ti
- Fix racing issue