Yes, it's as ludicrous as it sounds, but it's still true. Bizarrely,
the previous commit makes CLAHE run about 10% slower on Android, even
though it doesn't even touch any CLAHE code. Splitting it off fixes that,
although the reason it does is a mystery for the ages.

It's cleaner when it's in its own file, anyway. ;=]

This should allow many algorithms to take advantage of more parallelization
technologies.

For example, haarcascade_frontalface_alt2.xml is now supported.
Note that classifier's pattern of a cascade file must be consistent,
i.e., all trees must either have two nodes or one node, otherwise
unexpected results will occur.

Other fixes:
Test cases are updated.
Some unused codes are removed.
Fix some problems of haar when using OclCascadeClassifierBuf.

This fix is a workaround for current 2.4 branch without introducing an
additional oclMat buffer into CannyBuf object.
Test case is cleaned up.
Volatile keywords in kernels are removed for performance concern.

When argument pointers pushed into an vector and the pointers point to
address on stack, we need to make sure they are valid until kernels are
successfully flushed onto the queue.

Additional notes with this commit:
1. Add cornerHarris_dxdy and cornerMinEigenVal_dxdy to get
the interim dx and dy output of Sobel operator;
2. Add minMax_buf to allow user to reuse buffers in minMax;
3. Fix an error when either min or max pointer fed into minMax is NULL;
4. Corner sorter temporarily uses C++ STL's quick sort. A parallel
 selection sort in OpneCL is contained in the implementation but disabled
due to poor performance at the moment.
5. Accuracy test for ocl gfft.