How robust is journal rating in Humanities and Social Sciences? Evidence from a large-scale, multi-method exercise