Text this: Reinforcement Learning for Data-driven Workflows in Radio Interferometry. I. Principal Demonstration in Calibration