[feature][videostab] More options for the cv::videostab::KeypointBasedMotionEstimatorGpu

[Nuzhny007]

Just now I can't set some parameters for cv::videostab::KeypointBasedMotionEstimatorGpu class. For example I want to set non default parameters for features detector. On Nvidia Jetson Nano it works very slow. Whats about a more flexible interface? For example a second constructor with parameters from cuda::createGoodFeaturesToTrackDetector:

class CV_EXPORTS KeypointBasedMotionEstimatorGpu : public ImageMotionEstimatorBase

{

public:

    KeypointBasedMotionEstimatorGpu(Ptr<MotionEstimatorBase> estimator);

    KeypointBasedMotionEstimatorGpu(Ptr<MotionEstimatorBase> estimator,
                                    int maxCorners, double qualityLevel, double minDistance,
                                    int blockSize, bool useHarrisDetector, double harrisK);

...

[Nuzhny007]

Or like a CPU version above:

    void setDetector(Ptr<cuda::CornersDetector> val) { detector_ = val; }
    Ptr<cuda::CornersDetector> detector() const { return detector_; }