Sample Complexity of Goal-Conditioned Hierarchical Reinforcement Learning