FairSSD: Understanding Bias in Synthetic Speech Detectors