Peg monitoring L3: GraphQL oracle-events alert consumer

Context

Layer 3 (consumer side). Reads the Chainlink AnswerUpdated rows indexed by the Envio indexer (issue in chain-events/yearn-indexing-test) and turns anomalies into alerts. Mirrors the existing Envio→Telegram pattern.

Scope

• New script (mirror protocols/timelock/timelock_alerts.py) querying ENVIO_GRAPHQL_URL for recent rounds per feed.
• Detect: large round-over-round jumps, missed-heartbeat gaps, and sequence anomalies. Alert via utils/alert.py with per-asset channel routing from the shared registry.
• De-dupe across runs via the cache (same pattern as timelock_alerts.py).
• Add to the hourly (or 10-min) profile in automation/jobs.yaml.

Acceptance criteria

• Given indexed rounds, anomalies produce alerts; reruns do not re-alert on the same round.

References

• protocols/timelock/timelock_alerts.py, protocols/yearn/alert_large_flows.py (both use ENVIO_GRAPHQL_URL).
• Indexer issue: chain-events/yearn-indexing-test.
• Foundation issue (shared registry for routing).

Start by reading protocols/timelock/timelock_alerts.py and protocols/yearn/alert_large_flows.py, then inspect automation/jobs.yaml and the shared alert registry references. Use the indexed rounds from ENVIO_GRAPHQL_URL to verify anomaly detection and run the relevant monitoring profile; done means anomalies alert through utils/alert.py and reruns do not alert for the same round.