Draft: Support running multi-node

will resolve #2925

• libcluster
• cachex?
• shout: propagate initial state and every state change
  • need perfect replication, order matters
  • db?
  • we can generate an (flake)id upon every new message and resolve conflicts using id order
• config: propagate changes to all nodes
• restarter: outside Pleroma application tree, need other ways to communicate
• emoji: propagate changes to all nodes
• endpoint: should just work?
• oban: will just work https://elixirforum.com/t/example-of-multiple-node-worker-cluster-with-oban/29236/3
• streamer
• docker: allow only one active migration at a time (how?)
• ?