Removed offline Learning

Do not learn with offline data while collecting memories.
Prevents learning fram only a very few memories.

environment_wrapper.py

@@ -34,8 +34,6 @@ def one_episode(environment, agent, render, learn, max_steps=3000):
         if learn:
             agent.learn()
         if done:
-            if learn:
-                agent.learn(offline=True)
             break
     return score