Reinforcement Learning for Dynamic Resource Allocation in Optical Networks: Hype or Hope?