[Doc] Fix locked samples from RBs and ccl of tuto (#1132)

Author: vmoens

torchrl/data/replay_buffers/storages.py

@@ -276,7 +276,7 @@ def get(self, index: Union[int, Sequence[int], slice]) -> Any:
         out = self._storage[index]
         if is_tensor_collection(out):
             out = _reset_batch_size(out)
-        return out
+        return out.unlock_()
 
     def __len__(self):
         return self._len

tutorials/sphinx-tutorials/rb_tutorial.py

@@ -680,4 +680,12 @@ class MyData:
 # Conclusion
 # ----------
 #
+# We have seen how a replay buffer can be used in TorchRL, from its simplest
+# usage to more advanced ones where the data need to be transformed or stored
+# in particular ways.
+# You should now be able to:
+#
+# - Create a Replay Buffer, customize its storage, sampler and transforms;
+# - Choose the best storage type for your problem (list, memory or disk-based);
+# - Minimize the memory footprint of your buffer.
 #