Better memory allocation for Heuristic (#3785)

* Fixing some issues with previous action and better memory allocation in the Heuristic Policy

* Copying the data rather than using references

* [skip ci]Update com.unity.ml-agents/Runtime/Agent.cs

Co-Authored-By: Chris Elion <chris.elion@unity3d.com>

* Addressing comments

Co-authored-by: Chris Elion <chris.elion@unity3d.com>

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

             m_CumulativeReward = 0f;
             m_RequestAction = false;
             m_RequestDecision = false;
+            Array.Clear(m_Info.storedVectorActions, 0, m_Info.storedVectorActions.Length);
         }
 
         /// <summary>
                 return;
             }
 
-            m_Info.storedVectorActions = m_Action.vectorActions;
+            if (m_Info.done)
+            {
+                Array.Clear(m_Info.storedVectorActions, 0, m_Info.storedVectorActions.Length);
+            }
+            else
+            {
+                Array.Copy(m_Action.vectorActions, m_Info.storedVectorActions, m_Action.vectorActions.Length);
+            }
             m_ActionMasker.ResetMask();
             UpdateSensors();
             using (TimerStack.Instance.Scoped("CollectObservations"))
 
         void DecideAction()
         {
-            m_Action.vectorActions = m_Brain?.DecideAction();
+            }
+            var action = m_Brain?.DecideAction();
+
+            if (action == null)
+            {
+                Array.Clear(m_Action.vectorActions, 0, m_Action.vectorActions.Length);
+            }
+            else
+            {
+                Array.Copy(action, m_Action.vectorActions, action.Length);
             }
         }
     }

com.unity.ml-agents/Runtime/Policies/HeuristicPolicy.cs

         ActionGenerator m_Heuristic;
         float[] m_LastDecision;
         int m_numActions;
+        bool m_Done;
+        bool m_DecisionRequested;
 
         WriteAdapter m_WriteAdapter = new WriteAdapter();
         NullList m_NullList = new NullList();
         {
             m_Heuristic = heuristic;
             m_numActions = numActions;
+            m_LastDecision = new float[m_numActions];
         }
 
         /// <inheritdoc />
-            if (!info.done)
-            {
-                // Reset m_LastDecision each time.
-                 m_LastDecision = new float[m_numActions];
-                 m_Heuristic.Invoke(m_LastDecision);
-            }
+            m_Done = info.done;
+            m_DecisionRequested = true;
+
+            if (!m_Done && m_DecisionRequested)
+            {
+                 m_Heuristic.Invoke(m_LastDecision);
+            }
+            m_DecisionRequested = false;
             return m_LastDecision;
         }