Cloud deployment add agent health monitoring and auto restart on crash

[clintjeff2]

Motivation

• Cloud-deployed agents can fail silently; we need health visibility, error tracking, and a simple degraded-state trigger to surface problematic agents without blocking calls.
• Persisting errors atomically allows operators to inspect recent failures and count errors in a 24h window.

Description

• Add an atomic JSON-backed error store at lib/agents/agent-error-store.ts that records invocation errors to ./.data/agent-errors.json, exposes recordAgentExecutionError, recordAgentExecutionSuccess, recordAgentInvocation, getAgentErrorCount24h, and getAgentHealthSummary, and marks an agent degraded after 3 consecutive failures.
• Wrap cloud function and runtime agent executions to record invocations and errors (app/agent-functions/[id]/route.ts, lib/agent-runtime/agent.ts, lib/agent-runtime/cloud-agents.ts) so degraded agents remain callable, log a warning, and surface a degraded boolean.
• Add an authenticated admin health endpoint GET /api/admin/agents/[id]/health that returns { agentId, status, lastSeen, errorCount24h, degraded } (app/api/admin/agents/[id]/health/route.ts).
• Surface health info in the registry API/UI by returning errorCount24h and degraded from app/api/registry/route.ts and showing a degraded badge + Errors 24h in app/registry/page.tsx.
• Add tests for the error store behavior covering last-seen, 24h error counts, degraded transition after 3 consecutive errors, and that a success breaks the error streak (__tests__/agent-error-store.test.ts).

Testing

• Ran unit tests: npx vitest run __tests__/agent-error-store.test.ts __tests__/api/registry/registry.test.ts __tests__/api/agents/health.test.ts and all tests passed (3 files, 16 tests passed).
• Typecheck: npx tsc --noEmit reported failures unrelated to these changes due to missing optional dependencies (lru-cache, @wagmi/connectors) in the workspace.
• Full Next build: npm run build was attempted but blocked by external Google Fonts fetch failures during the Next.js build (network/font fetching issue) and not by the new code.

---

Closes #198

[sonarqubecloud]

Quality Gate passed

Issues
3 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[clintjeff2]

@leocagli, review and merge.

[clintjeff2]

@leocagli, review and merge.

[clintjeff2]

@leocagli, please review and merge.