Any Large Language Model Can Be a Reliable Judge: Debiasing with a Reasoning-based Bias Detector