Using cv::magnitude instead of custom loop and Mat<float> instead of Mat<double> to compute gradient magnitude channel.

added computeNMChannels helper function to compute the diferent channels to be processed independently in the N&M algorithm.

it conflicts with cv::detail

changed default parameters in CreateERFiltersNM{1,2} functions. They are not needed now in the sample example.

This allows to get rid of the last remaining Ptr<T>::addref calls.

Also changed the static functions to inline, to stop GCC complaining
about unused functions.

Now the features2d class wrappers use composition instead of inheritance.