[PyTorch] native_cpp_binding for size() and stride() (#49262)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/49262
This uses the newly-added native_cpp_binding feature to avoid
dispatcher overhead for `size()` and `stride()`.
ghstack-source-id: 118480533
Test Plan: CI
Reviewed By: bwasti
Differential Revision: D25446275
fbshipit-source-id: 1215eaa530d5aa3d501f89da8c99d0a487d8c1b6
