[FEATURE] Model Monitoring API — Serve live model metrics and predictions statistics

[gelluisaac]

Description

Create endpoints that serve model performance metrics and drift detection data
for the ModelMonitoringDashboard frontend component.

Endpoints

• GET /api/v1/monitoring/metrics — Latest model metrics (accuracy, F1, AUC, drift score)
• GET /api/v1/monitoring/performance-history?days=30 — Time series of metrics
• GET /api/v1/monitoring/drift-report — Feature drift analysis report
• GET /api/v1/monitoring/prediction-stats — Prediction volume and distribution stats
• GET /api/v1/monitoring/latency — API latency percentiles (p50, p95, p99)

Implementation Details

• Store metric snapshots in a ModelMetrics database table
• Periodically compute metrics from logged predictions (via background task)
• Drift detection using existing astroml/validation/ modules

Acceptance Criteria

• Metrics endpoint returns data compatible with ModelMonitoringDashboard.tsx
• Performance history returns at least 30 data points
• Latency endpoint exposes realistic percentile values
• Empty/not-yet-trained models return informative defaults, not errors

Labels

enhancement, api, monitoring, ml