Multi-Rationalization and Multi-Preferences