Avoid race condition in creating checkpoint directories (#530)

* Avoid race condition in creating checkpoint directories

In multi-GPU training, several processes run the code that creates checkpoint dirs. This fix avoids a probably rare situation (but it happened to me) where another process created a dir between the `exists` check and the `makedirs` call.

* Remove the now unneeded check for dir existence

pytorch_lightning/callbacks/pt_callbacks.py

@@ -198,8 +198,7 @@ class ModelCheckpoint(Callback):
         self.monitor = monitor
         self.verbose = verbose
         self.filepath = filepath
-        if not os.path.exists(filepath):
-            os.makedirs(filepath)
+        os.makedirs(filepath, exist_ok=True)
         self.save_top_k = save_top_k
         self.save_weights_only = save_weights_only
         self.period = period