Using Deep Reinforcement Learning for Application Relocation in Multi-Access Edge Computing