Use dynamic quantized linear partitioner of xnnpack (#2252)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/2252

For grupwise 4bit quant we need dynamic quantized linear partitioner. Ideally
-X option just uses both dqlinear as well as regular partitioner but the latter
doesnt yet work.
ghstack-source-id: 217594372

bypass-github-export-checks

Reviewed By: mikekgfb

Differential Revision: D54492109

fbshipit-source-id: 638f274dd2074818672aed738b361fc24927324c