feat: align nll_loss with PyTorch Python API

[voltjia]

Summary

• Expose one full nll_loss interface in PyTorch Python parameter order.
• Retain only the previous integer-reduction interface as a [[deprecated]] compatibility overload.

Motivation

The previous public interface exposed the lower-level ATen reduction enum and placed reduction before ignore_index. PyTorch users instead see size_average, ignore_index, reduce, and a string reduction in that order. The earlier revision added convenience overloads for every trailing default; this revision keeps the API surface minimal.

No standalone issue is associated with this alignment follow-up.

Type of Change

• feat - new feature / new operator / new platform
• fix - bug fix
• perf - performance improvement (no behavioral change)
• refactor - code restructuring without behavior change
• test - adding or fixing tests only
• docs - documentation only
• build / ci - build system or CI configuration
• chore - tooling, formatting, or other non-code changes
• Breaking change

Platforms Affected

• CPU (WITH_CPU)
• NVIDIA (WITH_NVIDIA)
• Iluvatar (WITH_ILUVATAR)
• MetaX (WITH_METAX)
• Cambricon (WITH_CAMBRICON)
• Moore (WITH_MOORE)
• Ascend (WITH_ASCEND)
• PyTorch C++ bindings (WITH_TORCH)
• Build system / CMake / CI
• Python bindings / user-facing API

Smoke Test Result

Host: ssh nvidia
Image: accelerator-dev/nvidia:latest
PyTorch: 2.10.0a0+b4e4ee8

cmake ... -DWITH_CPU=ON -DWITH_NVIDIA=ON -DWITH_TORCH=ON \
  -DINFINI_OPS_OPS=nll_loss -DINFINI_OPS_TORCH_OPS=nll_loss
cmake --build build-nvidia --target ops -j 16
Build passed; codegen emitted 1 ATen implementation overload.

python -m pytest tests/test_generate_wrappers.py -q
13 passed

python -m pytest tests/test_nll_loss.py --devices cpu nvidia -q
48 passed

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
CPU	Yes	targeted build passed	Included in 48 targeted cases
NVIDIA	Yes	targeted build passed	Included in 48 targeted cases
Iluvatar	No	N/A - not affected	N/A - not affected
MetaX	No	N/A - not affected	N/A - not affected
Cambricon	No	N/A - not affected	N/A - not affected
Moore	No	N/A - not affected	N/A - not affected
Ascend	No	N/A - not affected	N/A - not affected

Benchmark / Performance Impact

N/A. Reduction argument translation occurs during operator construction; the ATen execution path is unchanged.

Notes for Reviewers

API alignment

Interface	Signature	Alignment basis
InfiniOps primary	nll_loss(input, target, weight, size_average, ignore_index, reduce, reduction, out)	PyTorch names and Python order; explicit out remains last per CONTRIBUTING.md.
PyTorch public API	F.nll_loss(input, target, weight=None, size_average=None, ignore_index=-100, reduce=None, reduction="mean")	Pinned PyTorch source
Deprecated compatibility	nll_loss(input, target, weight, reduction, ignore_index, out)	Preserves the pre-alignment integer-reduction interface for migration.

• C++ callers spell PyTorch defaults explicitly; no prefix/default convenience overloads are added.
• size_average and reduce retain PyTorch's legacy override behavior when either is provided.
• Ruff 0.15.22 and clang-format 21.1.8 checks passed.