Address teardown issue

[Binyang2014]

Ignore cuda/cu errors during teardown. Some pointer may be invalid at this point

[Copilot]

Pull Request Overview

This PR addresses teardown issues by modifying error handling during cleanup operations. The changes improve robustness by ignoring certain CUDA/CU errors that may occur when pointers become invalid during teardown.

• Adds new CUDA error types to the teardown error detection functions
• Introduces a new macro to completely ignore CUDA driver API errors during cleanup
• Updates memory test to properly manage process group lifecycle

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
test/torch/memory_report.py	Improves process group management by using distinct variable names and explicitly destroying groups
src/gpu_utils.cc	Enhances error handling during teardown by expanding recognized teardown errors and adding ignore-only macro

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[caiomcbr]

LGTM

[caiomcbr]

LGTM