Gravar e-mail: Self-Supervised Spatiotemporal Representation Learning for Skeleton-Based Human Action Recognition