fix: update InfiniRT runtime API usage

[voltjia]

Summary

• Update InfiniOps' InfiniRT aliases to use infini::rt::runtime::Runtime and DeviceRuntime.
• Replace CUDA backend uses of MemcpyHostToDevice with kMemcpyHostToDevice.
• Add the Hygon runtime traits required by the normalized InfiniRT runtime API shape.

Motivation

This is the companion change for InfiniTensor/InfiniRT#11. InfiniRT now keeps CUDA Runtime API-shaped symbols under infini::rt::runtime, and normalized memcpy-kind constants use the kMemcpy... spelling. InfiniOps needs this update to build against the new public headers.

Closes #

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)
• Hygon
• PyTorch C++ bindings (WITH_TORCH)
• Build system / CMake / CI
• Python bindings / user-facing API

Smoke Test Result

# accelerator-dev/nvidia:latest, with companion InfiniRT branch installed
python3 -m pip install /workspace/InfiniOps --no-build-isolation \
  --config-settings=cmake.define.INFINI_RT_ROOT=/workspace/infini-rt-prefix \
  --config-settings=cmake.define.AUTO_DETECT_BACKENDS=OFF \
  --config-settings=cmake.define.WITH_TORCH=OFF \
  --config-settings=cmake.define.WITH_CPU=ON \
  --config-settings=cmake.define.WITH_NVIDIA=ON

Successfully built InfiniOps
Successfully installed InfiniOps-0.1.0

python3 -m pip install /workspace/InfiniOps --no-build-isolation --no-deps --force-reinstall \
  --config-settings=cmake.define.INFINI_RT_ROOT=/workspace/infini-rt-prefix \
  --config-settings=cmake.define.AUTO_DETECT_BACKENDS=OFF \
  --config-settings=cmake.define.WITH_TORCH=ON \
  --config-settings=cmake.define.WITH_CPU=ON \
  --config-settings=cmake.define.WITH_NVIDIA=ON \
  --config-settings=cmake.define.INFINI_OPS_SMOKE_BUILD=ON \
  --config-settings=cmake.define.INFINI_OPS_TORCH_OPS=abs,clamp,exp

Successfully built InfiniOps
Successfully installed InfiniOps-0.1.0

Test Results on Supported Platforms

Platform	Affected	Build / Smoke Result	Full Result / Notes
NVIDIA	yes	smoke passed	InfiniLM operator subset wheel built with companion InfiniRT branch; representative PyTorch wrapper smoke built for abs,clamp,exp.
Iluvatar	no	N/A	Not touched.
MetaX	no	N/A	Not touched.
Cambricon	no	N/A	Not touched.
Moore	no	N/A	Not touched.
Ascend	no	N/A	Not touched.
Hygon	yes	not run	Header-only runtime trait/name update; no Hygon hardware/toolkit in the validation environment.

Full `pytest` output (optional)

N/A

Benchmark / Performance Impact

N/A. This is an API compatibility update; no kernel implementation is changed.

Notes for Reviewers

• This PR depends on feat!: align runtime API and add runtime dispatch InfiniRT#11. CI jobs that build against current InfiniRT master may fail until that dependency is merged or the CI dependency is overridden.`n- Full InfiniCore/InfiniLM inference validation was attempted, but the remote environment repeatedly failed while downloading the xmake Boost dependency from GitHub. The failure happened before InfiniCore compilation and was network-related.
• The relevant compatibility checks for this PR passed: InfiniRT installed successfully, InfiniOps built against it, and representative PyTorch wrapper generation/build passed without doing the full wrapper set.