Learning Goal-Directed Object Pushing in Cluttered Scenes with Location-Based Attention