Deliver critical requests independently of Telegram #440

Epic: #437

Outcome

a CRITICAL incident has reliable emergency delivery even when Telegram is unavailable
or an earlier HIGH alert was dispatched.

Scope

• Persist emergency and notification delivery jobs independently for the same incident.
• Send signed requests with stable action IDs and retry transient HTTP failures.
• Distinguish durable acceptance from action success; track the executor's terminal result.
• Replace the shared HIGH/CRITICAL cooldown as an emergency-suppression mechanism with explicit
  incident/action deduplication. A HIGH dispatch cannot suppress escalation to CRITICAL.
• Keep alert history, routing, and the existing protocol allowlist as shared behavior rather
  than building an independent emergency path specifically for siUSD.

Acceptance criteria

Telegram failure, HIGH then CRITICAL, duplicate delivery, executor unavailability,
and sender restart are covered, and pending work/results remain visible and recoverable.

Dependencies

Coordinate the request/action ID and status contract with https://github.com/tapired/liquidity-monitoring/issues/452 before implementation; integration completion is blocked by that issue.

Related work

• Current sender: https://github.com/yearn/monitoring/blob/main/utils/dispatch.py
• Current Telegram-before-dispatch ordering: https://github.com/yearn/monitoring/blob/main/utils/alert.py

Include relevant behavior/failure tests and operational documentation in the implementation. Update monitoring metadata when monitor behavior changes.