Add VL-BERT to provide Image and Text shared vector space

[boba-and-beer]

Include a new VLBERT model, combining image and text into the same vector space.

https://github.com/jackroos/VL-BERT