In oder to guarantee SIMDness between the two compilation steps we choose the
highest SIMD size that is optimized in the kernel for the first step.

clip the found objects in HOG as well (http://code.opencv.org/issues/3825); added test to check CascadeClassifier on small images (http://code.opencv.org/issues/3710)

Conflicts:
	modules/objdetect/src/hog.cpp

fix android build

when true, use signed gradient for the angular histogram computation.
default to false for backward compatibility.

(cherry picked from commit 7ce11669)

1. someMatrix.data -> someMatrix.prt()
2. someMatrix.data + someMatrix.step * lineIndex -> someMatrix.ptr( lineIndex )
3. (SomeType*) someMatrix.data -> someMatrix.ptr<SomeType>()
4. someMatrix.data -> !someMatrix.empty() ( or !someMatrix.data -> someMatrix.empty() ) in logical expressions

Conflicts:
	modules/core/include/opencv2/core/cuda.hpp
	modules/cudacodec/src/thread.cpp
	modules/cudacodec/src/thread.hpp
	modules/superres/perf/perf_superres.cpp
	modules/superres/src/btv_l1_cuda.cpp
	modules/superres/src/optical_flow.cpp
	modules/videostab/src/global_motion.cpp
	modules/videostab/src/inpainting.cpp
	samples/cpp/stitching_detailed.cpp
	samples/cpp/videostab.cpp
	samples/gpu/stereo_multi.cpp

This allows to get rid of the last remaining Ptr<T>::addref calls.

Reverted back to the previous method where a public method groupRectangles was added to HOGDescriptor to take care of ROI and weight grouping

Also move cv::linemod to own header

Also move RTTIImpl class to the only usage

Made all STL usages explicit to be able automatically find all usages of
particular class or function.