ci(rate-limit): wait for etherpad readiness before running test (#7726)

* ci(rate-limit): wait for etherpad to be ready before running the test

The workflow starts the etherpad container in the background, then runs
`pnpm install`, then runs the test. On a warm pnpm-store the install can
finish before etherpad is listening on 9001, at which point nginx returns
502 for the test request and the run fails. Recent README-only commits
on develop hit this race on three consecutive runs.

Poll the nginx-proxied endpoint (port 8081 — also what the test uses)
until it stops returning 5xx, with a 2-minute timeout and `docker logs
etherpad-docker` on giving up to make diagnosis straightforward.

* ci(rate-limit): address Qodo review (nginx logs, tighter timeout)

- Name the nginx container so its logs can be captured when the readiness
  poll times out — previously nginx was started anonymously and a
  502 caused by nginx itself (rather than etherpad) would have been
  hard to diagnose from the workflow log alone.
- On timeout also dump `docker ps -a` for container-state visibility.
- Tighten the readiness wait: 30 iterations × (1s curl timeout + 1s
  sleep) gives ~60s budget instead of ~240s, which is still well above
  observed cold-start time and keeps the failure-fast contract.

Author: JohnMcLear

.github/workflows/rate-limit.yml

@@ -58,11 +58,36 @@ jobs:
         name: run docker images
         run: |
           docker run --name etherpad-docker -p 9000:9001 --rm --network ep_net --ip 172.23.42.2 -e 'TRUST_PROXY=true' epl-debian-slim &
-          docker run -p 8081:80 --rm --network ep_net --ip 172.23.42.1 -d nginx-latest
+          docker run --name nginx-docker -p 8081:80 --rm --network ep_net --ip 172.23.42.1 -d nginx-latest
           docker run --rm --network ep_net --ip 172.23.42.3 --name anotherip -dt anotherip
       -
         name: install dependencies and create symlink for ep_etherpad-lite
         run: pnpm install --frozen-lockfile
+      -
+        # The etherpad container is started in the background above; without
+        # this wait the test step can hit nginx before etherpad is listening
+        # and nginx returns 502, failing the run on a cold cache. Poll the
+        # nginx-proxied endpoint (which is also what the test hits) until it
+        # stops returning 5xx.
+        name: Wait for etherpad behind nginx to be ready
+        run: |
+          # ~60s budget: 30 iterations × (1s curl timeout + 1s sleep).
+          # Cold-start of the etherpad container is well under that.
+          for i in $(seq 1 30); do
+            if curl -fsS -o /dev/null --max-time 1 http://127.0.0.1:8081/; then
+              echo "etherpad is ready behind nginx"
+              exit 0
+            fi
+            sleep 1
+          done
+          echo "ERROR: etherpad behind nginx did not become ready in time"
+          echo "--- docker ps ---"
+          docker ps -a || true
+          echo "--- nginx-docker logs ---"
+          docker logs nginx-docker || true
+          echo "--- etherpad-docker logs ---"
+          docker logs etherpad-docker || true
+          exit 1
       -
         name: run rate limit test
         run: |