Update MultiheadAttention documentations (#20071)
Summary:
Add documentations to add_bias_kv, add_zero_attn, and attn_mask.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/20071
Differential Revision: D15213034
Pulled By: zhangguanheng66
fbshipit-source-id: c3db4b9e8527863420ba3ce6abf6098d3b0fb7a7
