Fix constant_fold cleanup for shared get_attr targets

[Kevin-Li-2025]

Summary

Fixes constant_fold cleanup when multiple get_attr nodes reference the same module attribute.

Before this change, constant_fold deleted the underlying GraphModule attribute as soon as it found one dead get_attr node. If another live get_attr node still referenced the same target, the graph became invalid and failed linting with:

Node shared_1 target shared references nonexistent attribute shared

This patch keeps the existing dead-node cleanup, but only deletes the underlying attribute when no other live get_attr node still uses the same target.

Test plan

• PYTHONDONTWRITEBYTECODE=1 PYTHONPATH=. /tmp/torchao-test-venv/bin/python -m unittest test.quantization.pt2e.test_quantize_pt2e.TestConstantFold.test_shared_get_attr_target_stays_live
• ruff check torchao/quantization/pt2e/constant_fold.py test/quantization/pt2e/test_quantize_pt2e.py
• git diff --check

Fixes #4420.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/4542

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.