fix push obs

com.unity.ml-agents/Runtime/Training/TrainingPolicy.cs

 
             if (m_HasLastObservation)
             {
-                m_buffer.Push(m_LastInfo, m_LastObservations, m_CurrentObservations);
+                m_buffer.Push(info, m_LastObservations, m_CurrentObservations);
             }
             else if (m_buffer.Count == 0)
             {
 
-            m_LastInfo = info;
+            m_HasLastObservation = !info.done;
+
-            }
-            m_HasLastObservation = true;
-
-            if (info.done == true)
-            {
-                m_buffer.Push(info, m_CurrentObservations, m_CurrentObservations); // dummy next_state
-                m_HasLastObservation = false;
             }
         }