Don't Overthink It: A Survey of Efficient R1-style Large Reasoning Models