Fix saving FlaubertTokenizer configs (#14991)

All specific tokenizer config properties must be passed to its base
class (XLMTokenizer) in order to be saved. This was not the case for
do_lowercase config. Thus it was not saved by save_pretrained() method
and saving and reloading the tokenizer changed its behaviour.

This commit fixes it.

src/transformers/models/flaubert/tokenization_flaubert.py

@@ -96,7 +96,7 @@ class FlaubertTokenizer(XLMTokenizer):
     max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
 
     def __init__(self, do_lowercase=False, **kwargs):
-        super().__init__(**kwargs)
+        super().__init__(do_lowercase=do_lowercase, **kwargs)
         self.do_lowercase = do_lowercase
         self.do_lowercase_and_remove_accent = False