Review for NeurIPS paper: SVGD as a kernelized Wasserstein gradient flow of the chi-squared divergence