Implementing Statistical Anomaly Detection in PromQL

Act as a Data Scientist operating within an SRE organization. Write a highly complex, mathematically dense PromQL query targeting [METRIC_NAME] (e.g., active HTTP requests) that acts as an autonomous anomaly detection engine. The query must mathematically trigger an alert exclusively when the current metric value violently deviates by more than [STD_DEV_THRESHOLD] standard deviations from its own moving average, calculated precisely over a massive [WINDOW_SIZE] (e.g., 1 week). Provide a deep technical explanation of the underlying calculus combining the `avg_over_time` and `stddev_over_time` functions in this context. Output the exact JSON code to configure a Grafana visualization that beautifully overlays the actual live metric directly on top of the calculated, shaded 'Normal Range' (Confidence Band) for visual context.