Add additional CUDA error handling macros (#74865)
Summary:
Introduces additional ways of handling CUDA errors that allow automated linters to detect if errors are being handled.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/74865
Test Plan: Sandcastle
Reviewed By: ngimel
Differential Revision: D35194530
fbshipit-source-id: f4fe61594edbfd81e97a4b605935961b893df167
(cherry picked from commit 919ddf677c5b9b46c5e493ed64346a5f2527bf08)
