Does collective efficacy moderate the associations between physical work load and intention to leave or sickness absence?