CI: Add native Windows ARM64 testing

[rwgk]

Description

Closes #2622.

Adds native Windows ARM64 GPU testing for the public 13.4.x workflow. The new
job uses one representative Python 3.13 / CUDA 13.4 configuration on the
windows-arm64-1xgb10 runner and consumes the existing public wheel artifacts.

The workflow now:

• publishes the compiled bindings/core test artifacts needed by the
  single-CUDA-major Windows ARM64 build;
• verifies ARM64 runner hardware, GPU access, native Python, and native compiled
  extension imports;
• runs pathfinder, bindings with both legacy and per-thread default streams,
  bindings Cython tests, and core tests; and
• includes the WoA build and test jobs in the final CI aggregation without
  weakening the existing event, owner, permission, or driver safeguards.

The WoA path remains prerelease-gated, matching the existing Windows ARM64
wheel-build policy. This draft is based on the continuation of #2588 and
therefore also carries the latest main-to-13.4.x merge; it should remain draft
while the first public WoA runner execution validates the experimental job.

Testing

• Changed-file pre-commit suite, including YAML validation and actionlint
• git diff --check
• Bash syntax validation for ci/tools/run-tests
• PowerShell parser validation for the Windows ARM64 bootstrap
• Independent workflow, artifact-contract, security, and private-prior-art reviews

Native WoA execution is intentionally left to the draft PR CI run.

Checklist

• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[rwgk]

/ok to test