Efficient Randomized Exploration In Reinforcement Learning: Bridging Theory And Practice