[quant] Add fp32/fp16 zero_point support for GPU fakeQuant (#65836)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/65836
Add a GPU implementation for GPU fakeQuant, similar to D30975238 (https://github.com/pytorch/pytorch/commit/60915eb8108960c05ef10e975d1ce1118b226a64)
ghstack-source-id: 139779416
Test Plan:
https://www.internalfb.com/intern/testinfra/testconsole/testrun/281475183488511/
{F667112564}
Reviewed By: b-koopman
Differential Revision: D31091679
fbshipit-source-id: 68fd483e6926c7fd565703c01d8ffb337b75dca5
