A Reinforcement Learning Model Of Gaze Following