OWSM v3.1: Better and Faster Open Whisper-Style Speech Models based on E-Branchformer