Reinforcement Learning Strategies for Compiler Optimization in High level Synthesis