Why historians (and everyone else) should care about counterfactuals