Should pick ports during replay of intermediate subsequences

[colin-scott]

We pick ports at the beginning of the MCSFinder simulation, but not during intermediate replays. This could interfere with (long-running) experiments, where another process could swoop in and take the port we chose at the beginning. This would then cause the MCS results to be highly inflated.