[Fix] Added dbias and dgelu kernels for ROCm

[AllenFarcas]

Description

Please include a brief summary of the changes, relevant motivation and context.

Fixes https://github.com/ROCm/frameworks-internal/issues/13667

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

• Added guards for getDeviceComputeCapability() which is for NVIDIA platforms
• Implementedfp8_quantize_rocm reusing CastVectorizedUnaryKernelLauncher and CastVectorizedUnaryGradKernelLauncher
• Implemented partial_reduce_kernel and reduce_dbias_rocm to efficiently reduce large sized inputs
• Enabled test_cast_dbias and test_cast_dbias_dgelu cpp tests for ROCm

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

[alextmagro]

mostly LGTM, just need to move some functions to rocm_cast_kernels.cuh for licensing. Great work!

[wenchenvincent]

@AllenFarcas There are some core sgpu test failures. Could you make sure that they're fixed?

[wenchenvincent]

LGTM.