[PyTorch] Explicitly specify quantized tensor usages needed for linear op backward

[timmoon10]

Description

I've experienced some failures with te.Sequential when the linear op recieves quantized tensors without the required usages. This PR is a quick bugfix that explicity specifies the required usages for the dgrad and wgrad GEMMs.

Type of change

• Documentation change (change only to the documentation, either a fix or a new content)
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Infra/Build change
• Code refactoring

Changes

• Explicity specify quantized tensor usages needed for linear op backward
• Include backward SwiGLU+cast fusion in tests (closes [PyTorch] Enable quantized activation backward kernels in operation-based API tests #1463)

Checklist:

• I have read and followed the contributing guidelines
• The functionality is complete
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

[timmoon10]

/te-ci pytorch