* accumulate, accumulateSquare, accumulateProduct and accumulateWeighted

Exp and Log IPP implementations are changed to hal interface;

consequences of changes

accuracy test

Solve issue 7063

  * The difference becomes too large when multiply is done in int16
  * To reproduce the test failure, IPP has to be switched off

Program flow is not easily apparent so hopefully ANY comments in the code could help new users.

"declare.in(img, WARMUP_RNG)" overrides input image with random values

Also added support for different input sizes (via cv::resize)

cudev (only for use in .cu files) & cudalegacy (only legacy code) not included

There is an issue with processing of abs(short) function for
negative argument.

Affected OpenCL devices:
- iGPU: Intel(R) HD Graphics 520 (OpenCL 2.0 )
- CPU: Intel(R) Core(TM) i5-6300U CPU @ 2.40GHz (OpenCL 2.0 (Build 10094))

GraphCutSeamFinderGpu requires cudalegacy.

  * avoid using vld1_f16 and vst1_f16 on gcc 4 series (Ubuntu 14.04)
  * guard correctly with #if
  * use static inline