Add histogram aggregation type

com.unity.ml-agents/Runtime/StatsRecorder.cs

         /// Values within the summary period are summed up before reporting.
         /// </summary>
         Sum = 2
+
+        /// <summary>
+        /// Values within the summary period are reported as a histogram.
+        /// </summary>
+        Histogram = 3
     }
 
     /// <summary>

ml-agents-envs/mlagents_envs/side_channel/stats_side_channel.py

     # Values within the summary period are summed up before reporting.
     SUM = 2
 
+    # All values within a summary period are reported as a histogram.
+    HISTOGRAM = 3
+
 
 StatList = List[Tuple[float, StatsAggregationMethod]]
 EnvironmentStats = Mapping[str, StatList]

ml-agents/mlagents/trainers/stats.py

             self.summary_writers[category].add_scalar(
                 f"{key}", value.aggregated_value, step
             )
-            if key == "Environment/Cumulative Reward":
+            if value.aggregation_method == StatsAggregationMethod.HISTOGRAM:
                 self.summary_writers[category].add_histogram(
                     f"{key}_hist", np.array(value.full_dist), step
                 )

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

 import abc
 import time
 import attr
+from mlagents_envs.side_channel.stats_side_channel import StatsAggregationMethod
+
 from mlagents.trainers.policy.checkpoint_manager import (
     ModelCheckpoint,
     ModelCheckpointManager,
         for name, rewards in self.collected_rewards.items():
             if name == "environment":
                 self.stats_reporter.add_stat(
-                    "Environment/Cumulative Reward", rewards.get(agent_id, 0)
+                    "Environment/Cumulative Reward",
+                    rewards.get(agent_id, 0),
+                    aggregation=StatsAggregationMethod.HISTOGRAM,
                 )
                 self.cumulative_returns_since_policy_update.append(
                     rewards.get(agent_id, 0)