Improved automatic thresholds in SQL Rules

We've made key improvements to the automatic threshold option in SQL Rules, giving users more accuracy and more control in detecting anomalies. The automatic threshold now supports:

• Low, medium, or high sensitivity. Users can adjust this through the UI or Monitors as Code, allowing for tighter or looser thresholds.
• Removing historical true positives from the machine learning, preserving the sensitivity of the thresholds. Incidents marked as fixed or helpful will be taken into account when training our ML model.
• Refactored model based on user feedback to improve accuracy.