adjust the header parser to support "public virtual" construction and make a tweak for multiple inheritance, fix potential memory problem with Python's IplImage.tostring() method (ticket #1486)

fixed bug #1531. added separate distanceTransformWithLabels function that returns labels, since normally the labels are not needed.

added the abstract cv::Algorithm class; low-level exp & fastAtan etc. functions for more efficient SIFT implementation; fixed bug #1521

Made changes in the stitching log macros: now the function stitchingLogLevel() may be used to make the stitching classes more/less verbose.

applied patch by Will Lucas; improved separable filter performance (in particular cv::GaussianBlur) on 16s images (for [much] faster SIFT)

e.g. "n += abs(sign(value))"  or  "n += sign(cv::norm(value) > eps_64"