Deep Reinforcement Learning-Based DRAM Equalizer Parameter Optimization Using Latent Representations