Successor Feature Landmarks for Long-Horizon Goal-Conditioned Reinforcement Learning Christopher Hoang 1