Add Docker cleanup strategy to prevent storage drift on Pi/server

[fe51]

Pitch

Each deploy creates a new container (old one is explicitly removed),
orphaning its logs and leaving dangling images from the local build.
On a Pi/server with limited storage this accumulates silently.

Technical suggestions

Two options to discuss:

• Deploy-time: add a docker system prune -f task in the engine role
  (or deploy-engines playbook), runs on each deploy
• Cron: add a periodic docker system prune -f in engine_cron role,
  decoupled from deploy

Note: avoid docker system prune -af in cron — would delete local tagged
images if container is temporarily stopped.