This patch adds ocl kernels to accelerate Dense Inverse Search
based optical flow algorithm, it acclerates 3 parts in the algorithm,
including 1) Structure tensor elements compute, 2) Patch inverse search,
3) Densification compute.

Perf and accuracy test are also added. The perf test shows it is 30%
faster than the current implementation.
Signed-off-by: Li Peng <peng.li@intel.com>

Added variational refinement as a separate class (based on implementation
inside DeepFlow, but significantly accelerated, about 4-6 times faster),
accelerated the main dense inverse search algorithm. Added several new
features including patch mean normalization for increased robustness to
illumination changes and spatial propagation, which often helps to recover
from errors introduced by the coarse-to-fine scheme. Expanded the
documentation, added new accuracy and perf tests. Refactored some of
the already existing optical flow accuracy tests.

DisparityWLSFilter demonstrated the best results, so I removed all the other
filters. Quality was significantly improved by adding confidence support
to the filter (left-right consistency + penalty for areas near depth discontinuities).
Filter was optimized using parallel_for_ and HAL intrinsics. Demo application was
rewritten for better compliance with OpenCV standards. Added accuracy and
performance tests. Documentation was added, as well as references to the
original papers.