[PyTorch] Migrate TI usage in ATen/native to borrowing (#58305)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/58305
Borrowing is more efficient, and we can see in all these cases that the TensorIterator doesn't outlive the input & output Tensors.
ghstack-source-id: 129598793
Test Plan: Existing CI
Reviewed By: ezyang
Differential Revision: D28445712
fbshipit-source-id: 0822f1408a0a71c8f8934e6d90659ae3baa085ac
