The case for mainstreaming gender in evaluation