feat(triton): add gemm operator

[fuyou4546]

Summary

• Add Gemm operator on Triton backend (src/triton/ops/gemm/build.py, gemm.py, gemm.h)

Motivation

Add a batched GEMM operator on the Triton backend, supporting fp16, bf16, and fp32 with alpha/beta scaling and optional transposition. The kernel uses a 2D tiled grid with GROUP_SIZE_M swizzle for L2 cache locality.

Closes N/A

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 (requires a ! in the Conventional Commits prefix or a BREAKING CHANGE: footer)

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

pytest tests -m smoke -q
······
73 passed, 13 skipped, 22328 deselected in 2.76s

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
NVIDIA	√	Successfully installed InfiniOps-0.1.0 / 73 passed, 13 skipped, 22328 deselected	1500 passed, 7500 deselected
Iluvatar
MetaX
Cambricon
Moore
Ascend

Full `pytest` output (optional)

pytest tests/test_gemm.py -k "cuda-8"
======= test session starts =======
platform linux -- Python 3.12.0, pytest-9.0.3, pluggy-1.6.0
rootdir: /home/zhangshuo/projects/InfiniTensor/InfiniOps
configfile: pyproject.toml
plugins: xdist-3.8.0, cov-7.1.0
collected 9000 items / 7500 deselected / 1500 selected
······
1500 passed, 7500 deselected in 5.35s

Benchmark / Performance Impact

N/A

Notes for Reviewers

Depends on the AOT infrastructure in feat/triton-backend.