[2026春季][T1-1-1] HyosungSink

[HyosungSink]

Summary

• Connect T1-1-1 operators to the InfiniCore Python use_ntops path: rad2deg, copysign, lcm, lgamma, nextafter.
• Keep the implementation at the Python dispatch layer; this PR does not add InfiniOP native kernels, pybind bindings, or C++ bridge implementations.
• Add Python-level tests that verify enabling infinicore.use_ntops dispatches to ntops.torch and matches PyTorch reference behavior.
• Include the Iluvatar validation build adjustment needed for this operator group.

Validation

• NVIDIA L4/sm_89 integration passed for all five operators through the use_ntops path.
• NVIDIA L4/sm_89 tests covered both out-of-place and out= execution for the connected operators.
• Iluvatar MR-V100 integration passed for all five operators through the use_ntops path.
• InfiniCore behavior remains scoped to Python dispatch; unsupported non-ntops paths are not changed by this PR.

Notes

• PyTorch is used only as the test reference.
• Runtime operator execution goes through the ntops.torch backend when use_ntops is enabled on supported GPU devices.
• Screenshots for the validation commands will be added before marking ready for review.