Adaptive k-Space Sampling in Magnetic Resonance Imaging Using Reinforcement Learning