How to Combat Reactive and Dynamic Jamming Attacks with Reinforcement Learning