Deep reinforcement learning for adaptive flexible job shop scheduling: coping with variability and uncertainty | Synapse