cosmetic changes to fix this warning.
real bugs not found

Fixed getSubset method to support Nx3 1-channel matrices as input (3xN 1-channel matrices doesn't supported at all).

Added test for checking that solvePnPRansac accepts vector<Point3f> and Nx3 1-channel or 1xN/Nx1 3-channel matrices as input.

Changed parameter order in documentation of findEssentialMat, because threshold and probability were switched.

Extension of the camera distortion model for tilted image sensors (Scheimpflug condition) including test

if they are not we also do not need to allocate a local copy of the rvec
and tvec

also replace C defines with C++ enums

use the same approach like in fisheye calibration: instead of setting
masked out rows to zero, remove them from the equation system.
This way JtJ does not become singular and we can use the much faster LU
decomposition instead of SVD.

This results in a speedup of the Calibrate unit tests of 3x-10x.

so it matches the code.

the first order approximation of the geometric error

Make a note about 16-bit signed format - the function assumes that
values have no fractional bits (so 16-bit disparity from StereoBM
and StereoSGBM cannot be directly used!)

improves performance by factor 1.2 to 2.0

Fix for https://github.com/Itseez/opencv/issues/4426
Documentation says that input points format is the same to input for findFundamentalMat

OPENCV_IPP env var now allows to select IPP architecture level for IPP9+;
IPP initialization logic was unified across modules;

IPP_VERSION_MAJOR * 100 + IPP_VERSION_MINOR*10 + IPP_VERSION_UPDATE
to manage changes between updates more easily.

IPP_DISABLE_BLOCK was added to ease tracking of disabled IPP functions;

Minor changes to IPP implementations;

Refer to the following issue for detail:
https://github.com/Itseez/opencv/issues/5145

`A || !A` is `true` so write `(A && B) || !A` as `B || !A`