Representation Learning for Event-based Visuomotor Policies