do not use the large "score" buffer; now without non-max suppression OpenCL FAST is pretty efficient

Made all STL usages explicit to be able automatically find all usages of
particular class or function.

possibly repaired FAST test (however, FAST<7/12> and <5/8> could still do something different than described in the paper)

added initModule_features2d(). Possibly solved problem when the linker excludes seemingly unused detectors/descriptors.

lot's of changes; nonfree & photo modules added; SIFT & SURF -> nonfree module; Inpainting -> photo; refactored features2d (ORB is still failing tests), optimized brute-force matcher and made it non-template.

minor build system changes (now cuda code in opencv_core is compiled using CUDA_ARCH* cmake variables)

added improved ORB implementation, convex-convex polygon intersection, eigen2x2 low-level function ...

updated 3rd party libs: CLapack 3.1.1.1 => 3.2.1, zlib 1.2.3 => 1.2.5, libpng 1.2.x => 1.4.3, libtiff 3.7.x => 3.9.4. fixed many 64-bit related VS2010 warnings