Can care staff accurately assess health-related quality of life of care home residents? A secondary analysis of data from the OPERA trial