Multi-armed Bandit Experimental Design: Online Decision-making and Adaptive Inference