Shape, Pose, and Appearance from a Single Image via Bootstrapped Radiance Field Inversion