A data science team runs analysis workflows on large tabular datasets using pandas, but iteration speed is limited by CPU processing. They want to accelerate their existing pandas code on GPUs without rewriting their scripts. Which CUDA-X library is designed to meet this need?