Standard model v2.0: How visual cortex might learn a universal dictionary of shape components