Resolves #228, resolves #229, resolves #230, resolves #231

[JONAH-6]

Summary

Resolves #228, resolves #229, resolves #230, resolves #231

---

What was done

#228 — Pagination for stream events list

• GET /v1/streams/{id}/events now returns { data, total, hasMore } instead of a bare array
• Supports limit (default 50, max 500), offset, cursor=<eventId>, and direction=asc|desc
• Cursor-based pagination uses Prisma's native cursor API (skip-the-cursor, stable ordering by
  createdAt)
• Added @@index([createdAt]) and @@index([streamId, createdAt]) composite index to StreamEvent in
  prisma/schema.prisma
• Updated Swagger docs for the events endpoint

#229 — Graceful shutdown with SSE connection draining

• On SIGTERM / SIGINT: sets a shutdown flag → sends event: reconnect to all connected SSE clients →
  stops accepting new SSE connections (503) → stops the HTTP server → waits up to 30 s for the
  in-flight indexer batch → disconnects Redis → disconnects Prisma → exits 0 (1 on timeout)
• SorobanEventWorker.waitForDrain() tracks the active batch promise and resolves when idle
• sseService.isShuttingDown() checked in the SSE subscribe handler

#230 — Redis pub/sub for horizontal SSE scaling

• New backend/src/lib/redis.ts: initializes a publisher + subscriber pair using ioredis, with a
  retryStrategy and graceful fallback — if REDIS_URL is unset or Redis is unreachable, the service
  runs in single-instance in-process mode automatically
• sseService.broadcastToStream / broadcastToUser: when Redis is available, publishes to
  sse:stream:{id} / sse:user:{key}; otherwise falls back to local broadcast
• Each instance PSUBSCRIBEs to sse:stream:* and sse:user:* on startup and routes incoming messages
  to local SSE clients
• REDIS_URL added to .env.example with documentation
• ioredis ^5.3.2 added to package.json

#231 — Admin metrics endpoint

• New backend/src/routes/adminRoutes.ts mounted at /v1/admin
• GET /v1/admin/metrics — requires Authorization: Bearer <ADMIN_SECRET>; returns stream counts by
  status, events indexed in last 24 h, SSE connection count, indexer last ledger + lag seconds, and
  server uptime
• GET /health updated to include db, indexer.status, and indexer.lastLedger fields; still returns
  503 when DB is down
• ADMIN_SECRET added to .env.example