Apply noexcept to relevant move methods to improve performance (#92156)
This clang-tidy check is disabled globally due to false positives on containers, but there are a few places here where adding clang-tidy would actually improve performance (by allowing STL containers to use the move operator / assignment)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/92156
Approved by: https://github.com/ngimel
