Reinforcement Learning from Implicit Neural Feedback for Human-Aligned Robot Control