End-to-end performance optimization placeholder issue

[bxyu-nvidia]

We will be doing a performance analysis across the end-to-end Gym + RL pipeline. For any low-hanging inefficiencies we would like to address them. This is a placeholder issue for the inefficiencies we may identify.