Text this: Pose‐guided adversarial video prediction for image‐to‐video person re‐identification