Fixed bug in multiprocess environment

- missing copy of internal state before returning the current state
- now vectorized pensulum works

mushroom_rl/core/multiprocess_environment.py

@@ -129,7 +129,7 @@ def step_all(self, env_mask, action):
             else:
                 step_infos.append({})
 
-        return self._states, rewards, absorbings, step_infos
+        return self._states.copy(), rewards, absorbings, step_infos
 
     def render_all(self, env_mask, record=False):
         for i, remote in enumerate(self._remotes):