Reliable Explanations or Random Noise? A Reliability Metric for XAI