allows using Mat and vector<Vec3d> as argument for rvecs, tvecs that are
allocated at once.

opencv_dnn: replaced static LayerFactory::impl member with a static function that constructs the object on the first use (fix for bug #383)

This way, we are certain not to depend on the order of static
variable initialization (LayerFactory::impl in dnn.cpp, vs. init
in init.cpp), which is causing bug #383.

also avoids manual memcpy: the predefined dictionary data is not
duplicated any more.

store marker internally as
"r0_b0, r0_b1, .. , r0_bn, r2_b0, .." instead of
"r0_b0, r1_b0, .. , rn_b0, r1_b0, .."
i.e. the same as in predefined_dictionaries.cpp.
This makes loading a dictionary a simple memcpy and also allows to get
rid of the custom hammingDistance implementation in favor of
hal::normHamming.

* Sparse match interpolator interface and EdgeAwareInterpolator were
added to the ximgproc module
* New optical flow algorithm, based on PyrLK sparse OF and sparse match
interpolation, is added to the optflow module