fix: dockerize backend reliably and optimization

[AlexKalll]

What & Why

The backend couldn't run reliably in Docker: it crashed on startup,
demo users couldn't be seeded, and login from the frontend failed.
This PR fixes all three and slims down the build.

Fixes

• App crash on startup - pinned @nestjs/swagger to 11.2.6
  (11.4.x breaks nestjs-zod → ERR_PACKAGE_PATH_NOT_EXPORTED).
• Seeding in production - added seed:prod script that runs the
  compiled seeders from dist/ (plain seed needs ts-node, which is
  pruned from the production image). Seed is now idempotent + logs the
  created count; added a hint when ENABLE_DEMO_USERS_SEED is off.
• Stale role - removed the unused Admin role from the role seed.
• Regenerated package-lock.json with npm 11 (lockfileVersion 3).

Optimization (build time ~9 min → ~3s)

• Dockerfile pins npm to 11.5.2 (matches lockfile).
• Single npm ci in the builder stage + BuildKit cache mounts for the
  npm cache and the TypeScript incremental cache → near-instant rebuilds.
• npm prune --omit=dev in the builder; pruned node_modules is copied
  into the runtime stage (no second full install).
• app service: added healthcheck, init: true, and log-rotation
  limits.

How to test

cd <path-to-hyperdata-backend>
docker compose down -v --rmi all
docker compose build app
docker compose up -d
docker compose exec app npm run seed:prod

# login: via frontend