Evidence for a modest undercount bias in early historical Atlantic tropical cyclone counts