Add dcccpy platform runtimes and Linux ARM64 release builds

[tctco]

Summary

• add the dcccpy Python wrapper with slim and platform runtime installation paths
• package Linux x86_64, Windows x86_64, macOS arm64, and Linux arm64 runtimes
• build Linux arm64 natively on GitHub Actions using ubuntu-24.04-arm and a manylinux2014 aarch64 container
• make dcccpy[runtime] select the matching runtime package through platform markers
• add macOS Gatekeeper guidance for both dcccpy and Slicer execution failures
• attach the Linux arm64 runtime wheel and dcccpy distributions to tagged GitHub releases

Why

NVIDIA DGX Spark requires an aarch64 DCCCcore executable. The existing release workflow only produced Linux x86_64 builds, and dcccpy had no Linux arm64 runtime package for pip to resolve.

User impact

After the arm64 runtime and dcccpy distributions are published to PyPI, supported Linux arm64 users can install the native executable with:

python -m pip install -U "dcccpy[runtime]"

The plain dcccpy package can also download DCCCcore-<version>-ubuntu-latest-arm64.zip from GitHub Releases on first use.

Validation

• dcccpy tests: 11 passed, 1 skipped
• Slicer macOS hint tests: 4 passed
• built dcccpy 0.1.6 wheel successfully
• verified runtime wheel tag: py3-none-manylinux2014_aarch64
• twine check passed for dcccpy and the arm64 runtime wheel
• workflow YAML parsing and git diff --check passed

The final native aarch64 compile and executable smoke test are intentionally delegated to the GitHub-hosted ARM64 runner.