Slice 6: Detect anomalies and populate the issue feed

[newbish]

Objective

Turn waste and breakage into visible operating signals.

Tasks

• Define anomaly rules for spend spikes, retry spikes, latency spikes, and cache drops
• Generate issue_events during rollup
• Display severity and trend context in the dashboard
• Add basic threshold configuration and docs

Acceptance criteria

• Seeded anomaly scenarios appear in the issue feed
• Threshold behavior is predictable and documented
• Severity levels are visible and useful, not just colorful