Path-by-path optimal control of switched and Markovian jump linear systems