Fix: crmd: All peers need to track node shutdown requests

This allows them to have the correct value for 'expected status' if they ever
become elected as DC at a point when multiple nodes died.