V1 Recovery R11 — Complete Operations integration, alerts, and notifications

[tonywxx]

Part of #129.

Problem Statement

Operations foundations are not fully wired to real Data, Paper, Bot, and Feedback failures or User notifications.

Solution

Project append-only operational evidence into health, alerts, safety actions, notifications, and drill-down views.

Acceptance Criteria

• Data, worker, model, account, Risk/OMS, adapter, local-system, and feedback health are independently visible.
• Typed alerts support Active/Acknowledged/Resolved lifecycle, debounce/hysteresis, redaction, and retained evidence.
• Required unhealthy/unknown states trigger their frozen fail-closed safety actions.
• Notification Center, critical banner, OS notifications, and drill-down work in both locales without giving frontend state authority.

Out of Scope

Cloud observability, telemetry collection beyond declared local bounds, or automated research changes.

Planning Source

• docs/v1-completion-recovery-map.md
• docs/adr/0090-rebaseline-v1-to-okx-spot-end-to-end.md

Implementation must run in one independent $implement <issue> session. Complete every acceptance criterion, record exact verification in an English issue comment, and close only after all evidence passes.

[tonywxx]

This issue is retained as a future Operational Observability and Monitoring candidate under the fifteen-gate delivery sequence specified by #144.

It is not an executable frontier now. Its scope and current-head gaps must be revalidated when Gate 14 is explicitly User-authorized. Historical implementation and evidence remain useful, but do not grant Module Acceptance or Workflow Continuation Approval.