Phase 3.1 -- Analytics Trends & Forecast

Parent: #762

Context

The new dashboard needs sparkline trend data and budget forecast projections. The backend has the raw data (CostRecord, TaskMetricRecord) but no time-series aggregation endpoints.

Scope

GET /api/v1/analytics/trends

• Query params: period (7d/30d/90d), metric (tasks_completed, spend, active_agents, success_rate)
• Response: array of {timestamp, value} datapoints bucketed by day (or hour for 7d)
• Data sources: CostRecord (spend), TaskMetricRecord (tasks, success rate), agent status snapshots

GET /api/v1/analytics/forecast

• Query params: horizon_days (7/14/30)
• Response: {projected_total, daily_projections: [{day, value}], days_until_budget_exhausted, confidence}
• Data source: CostTracker's internal trending logic (needs to be exposed as public method)

Extend GET /api/v1/analytics/overview

• Add fields: budget_remaining_usd, budget_used_percent, cost_7d_trend (sparkline array), active_agents_count, idle_agents_count

Deliverables

• Trends endpoint with 4 metric types
• Forecast endpoint with daily projections
• Extended overview response
• Unit tests with realistic test data
• OpenAPI schema updated

Notes

• Can start immediately -- no dependency on frontend work
• Expose CostTracker's trend logic via new public methods, don't duplicate