Use bert init for xlm_base

Summary:
Use bert init for xlm_base.  This seems to be much closer to what is done in the [XLM](https://github.com/facebookresearch/XLM/blob/master/src/model/transformer.py#L44) repo.

At update 10 with BERT init (f121471600), loss starts at 14.234

At update 10 without BERT init (f121471612), loss starts at 154.423

Reviewed By: liezl200, pipibjc

Differential Revision: D15874836

fbshipit-source-id: f81bf83a078992d7476ba7fdf263b731a9f5b66d

fairseq/models/masked_lm.py

@@ -352,4 +352,5 @@ def xlm_architecture(args):
     args.activation_fn = getattr(args, 'activation_fn', 'gelu')
     args.encoder_normalize_before = getattr(args, 'encoder_normalize_before', False)
     args.pooler_activation_fn = getattr(args, 'pooler_activation_fn', 'tanh')
+    args.apply_bert_init = getattr(args, 'apply_bert_init', True)
     base_architecture(args)