correct range computations, it is difficult to do the more efficient single pass 3N/2 implementations correctly in this particular setting without costing more than they save

Most importantly, added support for multiple include directores,
but also did some general cleanup.

* use explicit extrapolation only for sub-matrixes
* use built-in interpolation for INTER_NEAREST mode

Changed the brief description of function Mat::colRange from "Create a matrix header for the specified row span." to "Create a matrix header for the specified column span."

On a 32-bit compiler the calculation may result in data (size_t) overflow when running some paralleled algorithms (which can safely run on a 64-bit compiler).
This bug is found when running OpenCV's Retina tutorial on 32bit VS2010.

They're not listed in the documentation for the FindZLIB module,
so we shouldn't use them unless necessary.

don't calc diff term if it is not used for epsilon criterion

Correct reading PNG color type palette with or without alpha imread flags -1 or 1
Better not using pnginfo.h, using png_get_tRNS instead

incorrect types for template instantiation