Scaling Laws for Discriminative Speech Recognition Rescoring Models