fix: target the dokploy container in the auth secret migration script

[Hayyan612]

Fixes Dokploy/dokploy#5359

The 2FA migration step in security/0.29.3.sh finds the Dokploy container with:

docker ps --filter "name=dokploy" --format "{{.ID}}" | head -n1

Docker's name filter is a substring match, so this also matches dokploy-postgres and dokploy-traefik, and docker ps lists the newest container first. When Postgres or Traefik was started after Dokploy, for example after a reboot, the migration ran inside the wrong container and failed with can't cd to /app.

This selects the task by the label Swarm sets on every service task, com.docker.swarm.service.name=dokploy, as suggested in the issue. Label filters compare the value exactly, so only the dokploy service's container matches.

Testing

Ran the original and the fixed script against a local Swarm laid out like install.sh: a dokploy service, a standalone dokploy-traefik container, and a dokploy-postgres service created last so it is listed first. id and timeout were shimmed to run on macOS.

original: docker ps order: dokploy-postgres.1.… dokploy-traefik dokploy.1.…
          sh: cd: line 0: can't cd to /app: No such file or directory
          Error: 2FA migration failed (exit code 2)        exit 1

fixed:    docker ps order: dokploy-postgres.1.… dokploy-traefik dokploy.1.…
          ✅ 2FA records migrated
          🔄 Updating Dokploy service...                   exit 0