Add doc warnings for default SELU gain (#54057)
Summary:
Fixes https://github.com/pytorch/pytorch/issues/24991 and provides the alternative solution suggested in https://github.com/pytorch/pytorch/issues/53694. Also related to https://github.com/pytorch/pytorch/issues/54055
Attempt to make people aware of the difference between paper and implementation of SELU gain.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/54057
Reviewed By: ailzhang
Differential Revision: D27292060
Pulled By: jbschlosser
fbshipit-source-id: e0e303595e6a7d05d11dfb68735e1839f55987a2
