S3RL: Enhancing Spatial Single‐Cell Transcriptomics With Separable Representation Learning | Synapse