fix: harden SSR server with keepalive fix, error handling, and health check

[jeanpaulsio]

Summary

Ports the SSR hardening from step-thru (PR #125) to this template.

Root cause of overnight 502s on Render: Node's default `keepAliveTimeout` (5s) is shorter than Cloudflare's idle timeout (~100-900s). After 5s idle, Node closes the TCP connection. Cloudflare doesn't know, sends a request on the dead socket, gets a TCP RST, returns 502. The failure happens at the TCP layer, invisible to Express/Sentry/application logging.

Fix: set `server.keepAliveTimeout = 120000` and `server.headersTimeout = 125000` (must be greater than keepAliveTimeout per Node.js requirement). See Render docs and nodejs/node#59193.

Additional hardening

• try/catch around `renderPage()` with 500 fallback HTML page
• 30s request timeout middleware (kills stalled SSR renders)
• `GET /health` endpoint + `healthCheckPath` in `render.yaml`
• Global Express error handler as safety net
• Slow request logging (>5s) and 5xx error logging (structured JSON)
• Process lifecycle logging (SIGTERM, uncaughtException, unhandledRejection)
• Memory monitoring (warns at 400MB heap, checked every 60s)
• Startup log with app name, port, and Node version

Test plan

• `cd web && npm ci && npm run build` succeeds
• `cd web && node server.js` starts and logs structured JSON
• `curl localhost:3000/health` returns `{"status":"ok"}`
• App-generated repo deploys to Render without 502s overnight