[BugFix] correct the use of step_mdp method in data collector (#637)

* rectify step_mdp usage

* run linting

* run linting

Co-authored-by: vmoens <vincentmoens@gmail.com>

Author: adityagandhamal

torchrl/collectors/collectors.py

@@ -633,12 +633,7 @@ def rollout(self) -> TensorDictBase:
                 tensordict_out.append(self._tensordict.clone())
 
                 self._reset_if_necessary()
-                self._tensordict.update(
-                    step_mdp(
-                        self._tensordict.exclude("reward", "done"), keep_other=True
-                    ),
-                    inplace=True,
-                )
+                self._tensordict.update(step_mdp(self._tensordict), inplace=True)
             if self.return_in_place and len(self._tensordict_out.keys()) > 0:
                 tensordict_out = torch.stack(tensordict_out, len(self.env.batch_size))
                 tensordict_out = tensordict_out.select(*self._tensordict_out.keys())