llama.cpp b10051 fixes SME vs SME2 kernel dispatch distinction

The llama.cpp project released version b10051, which addresses a critical issue in the integration of Scalable Matrix Extension (SME) capabilities. The update introduces a build-time and runtime distinction between SME(v1) and SME2 to ensure correct kernel dispatch.

• Previous versions treated SME as a single capability, incorrectly dispatching SME2-specific instructions on hardware that only supports SME(v1).
• The fix wires SME(v1) and SME2 kernels based on actual hardware support, preventing incorrect execution on SME(v1)-only systems.
• Binaries are available for macOS (Apple Silicon and Intel), Linux (x64, arm64, s390x, Vulkan, ROCm, OpenVINO, SYCL), Android, Windows (CPU, CUDA 12/13, Vulkan, OpenCL, HIP, OpenVINO, SYCL), and openEuler.

This change ensures that llama.cpp runs correctly on a wider range of ARM hardware by respecting the specific SME version supported by the CPU.