Random Path Selection for Continual Learning