Excluding the Irrelevant: Focusing Reinforcement Learning through Continuous Action Masking