NVIDIA Advances Vera Rubin Inference With New LPX and CPX Platforms for Faster AI Performance, Lower Token Costs