[Feature] Set up remote GPU runners for CUDA and ROCm regression testing

Is your feature request related to a problem? Please describe.

Currently, MCDC GPU regression tests are limited by local execution environments. This makes it difficult to consistently run GPU-based CI tests across contributors and development environments.

GPU support is currently undergoing active refactoring and is not fully operational yet. Setting up remote runners for both CUDA and ROCm backends would help future automated validation and regression testing.

Describe the feature you'd like.

Set up the remote runners for both CUDA and ROCm backends for MCDC GPU regression tests. The goal is to enable automated GPU regression testing through CI workflows using our remote GPU resources.

Additional context.

This setup will improve CI coverage for GPU-related development and help catch regressions earlier in the development cycle.

How can this issue be closed?

• Remote GPU runners are configured and accessible
• GitHub Actions workflows successfully trigger GPU regression tests
• GPU regression tests run automatically in CI