when dropping optimizations in the last commit, I forgot to keep the
simplest case where a single memset can be called

Only keep the ICV_HLINE_X optimization to reduce code size.

added ICV_HLINE custom implementations for element sizes up to 32
but timings show that it is not very relevant for sizes >= 12

added 64b optimization for 3 channels case
not added 64b optimization for 4 channels case since timings did not
show any improvement
split ICV_HLINE cases into inline functions instead of macro for code
size reduction, without significand speed drawback at first sight

# Conflicts:
#	modules/imgproc/src/drawing.cpp

On MacOS and iOS, the unused opencvBigToHost32 is a warning for buildbot

ICV_HLINE is split into several specific cases, according to pixel_size,
to optimize memory copies of the same color components along the line.

cv2.findContours() no longer modifies source image (#8150)

* cv2.findContours() no longer modifies source image

Since OpenCV 3.2, cv2.findContours() no longer modifies the source image but returns a modified image as the first of three return parameters.  ??? Do I have that correct ???

Also fixed up the code blocks to be valid markdown.

* ```python --> @code{.py}

Enables syntax highlighting in docs.

Solve issue 8136 opencv doc of threshold

Add a link to MatExpr in Detailed Description of Mat

The backslash causes build failure when cross-compiling with mingw-w64

fix for opencv/opencv#8105, compilation issue with mingw32 (in
google/googletest#721 a similar issue was solved and the reason was
described as MinGW defines _CRITICAL_SECTION and _RTL_CRITICAL_SECTION
as two separate (equivalent) structs, instead of using typedef)

CoInitialize/CoUninitialize in VFW classes (#8112)