Use resamp policy for SAC

ml-agents/mlagents/trainers/sac/trainer.py

             self.is_training,
             self.load,
             tanh_squash=True,
+            resample=True,
         )
         for _reward_signal in policy.reward_signals.keys():
             self.collected_rewards[_reward_signal] = defaultdict(lambda: 0)