Replace poseTrack.py/removeDoubleCount with nonMaximaSupression.

[windsor718]

Confidence will be calculated from poseNet results.

[windsor718]

def nonMaximaSupression(self, boxes, confidences):
        """
        supress the double counts based on the non maxima supression.

        Args:
            boxes (list): list of bounding boxes
            confidences (list): list of confidences

        Nots:
            the order of boxes and confidences must be same.
        """
        idxs = cv2.dnn.NMSBoxes(boxes, confidences, self.confThsld, self.nonMaximaSupThsld)
        return idxs