Assigning time-linked exposure status to controls in unmatched case-control studies: alcohol use and nearly lethal suicide attempts