Use native GPU architecture when NVIDIA GPU is detected; otherwise fall back to multi-arch build.

[mahdiehghazim]

This change makes MSCCL++ automatically select CUDA architectures based on the build environment. If an NVIDIA GPU is detected, the build targets the native GPU architecture for optimal performance; otherwise, it falls back to building for multiple architectures for portability. When building for the native architecture, FP8 support is automatically enabled for “a-series” GPUs (e.g., sm_100a), allowing the appropriate optimized code paths to be picked up.

[Copilot]

Pull request overview

This PR implements automatic CUDA architecture selection for MSCCL++ builds. When an NVIDIA GPU is detected at build time, the build targets the native GPU architecture for optimal performance; otherwise it falls back to building for multiple architectures (80, 90, 100, 120 based on CUDA toolkit version) for portability. The native architecture build enables automatic FP8 support for "a-series" GPUs.

Changes:

• Modified architecture selection logic to use "native" when NVIDIA GPU is detected
• Preserved multi-architecture fallback when no GPU is present