Hierarchical Reinforcement Learning with Model-Based Planning for Finding Sparse Rewards