Curriculum Reinforcement Learning using Optimal Transport via Gradual Domain Adaptation