Add zero_mask() for Vec256<BFloat16> (#37114)
Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/37114
Test Plan: Imported from OSS
Differential Revision: D21351861
Pulled By: VitalyFedyunin
fbshipit-source-id: 4564624cb33555a3f026af25540b2df24edaecfb