“Some,” and possibly all, scalar inferences are not delayed: Evidence for immediate pragmatic enrichment