llama.cpp b9964 release prevents duplicate spec model downloads

The llama.cpp project has published the b9964 release, which includes a fix to prevent duplicate specification model downloads.

• The source code change is tracked in pull request #25527.
• macOS Apple Silicon builds with KleidiAI are disabled in this release.
• openEuler support for standard x86 builds is disabled, but specific 310p and 910b (ACL Graph) variants remain available.
• Binaries are provided for macOS (Apple Silicon and Intel), iOS, Linux (Ubuntu with CPU, Vulkan, ROCm 7.2, OpenVINO, SYCL), Android, Windows (CPU, CUDA 12/13, Vulkan, OpenVINO, SYCL, HIP), and openEuler.
• A standalone UI package is also included.