Enable quantization for mobilebert (#477)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/477

Quantized: https://www.internalfb.com/intern/everpaste/?handle=GDQbkxamK5wUoYUDALGrDOPik_tUbsIXAAAB

didn't check against fx since the model is not symbolically traceable

Reviewed By: cccclai

Differential Revision: D49379415

fbshipit-source-id: 57c6361120d1bea6c58f3dad806eb57f3fef3b29