dynamic quantize linear and lower to xnnpack (#1849)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/1849

Add a path to dynamic quantize linear layers along with embedding. Will factor our embedding quant in a separate diff.
ghstack-source-id: 214288401
exported-using-ghexport

Reviewed By: digantdesai

Differential Revision: D53316360

fbshipit-source-id: 39d6a25d9da44ba84995c5d54a4a132da262b35d