Remove some vestigial code

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

         The reward signal generators must be updated in this method at their own pace.
         """
         buffer_length = self.update_buffer.num_experiences
-        self.cumulative_returns_since_policy_update.clear()
 
         # Make sure batch_size is a multiple of sequence length. During training, we
         # will need to reshape the data into a batch_size x sequence_length tensor.

ml-agents/mlagents/trainers/rl_trainer.py

     def __init__(self, *args, **kwargs):
         super(RLTrainer, self).__init__(*args, **kwargs)
         self.param_keys: List[str] = []
-        self.cumulative_returns_since_policy_update: List[float] = []
         self.step: int = 0
         self.training_start_time = time.time()
         self.summary_freq = self.trainer_parameters["summary_freq"]
         A signal that the Episode has ended. The buffer must be reset.
         Get only called when the academy resets.
         """
-        for agent_id in self.episode_steps:
-            self.episode_steps[agent_id] = 0
-        self.episode_steps[agent_id] = 0
-                self.cumulative_returns_since_policy_update.append(
-                    rewards.get(agent_id, 0)
-                )
                 self.reward_buffer.appendleft(rewards.get(agent_id, 0))
                 rewards[agent_id] = 0
             else:

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

         N times, then the reward signals are updated N times, then reward_signal_updates_per_train
         is greater than 1 and the reward signals are not updated in parallel.
         """
-
-        self.cumulative_returns_since_policy_update.clear()
         n_sequences = max(
             int(self.trainer_parameters["batch_size"] / self.policy.sequence_length), 1
         )