llama.cpp b9810 release adds cublasSgemmBatched mapping and new binaries

The llama.cpp project has released version b9810, introducing a CUDA mapping for `cublasSgemmBatched` in HIP/MUSA vendor headers. This update is accompanied by a comprehensive set of pre-built binaries for macOS, Linux, Windows, Android, and openEuler platforms.

• Added cublasSgemmBatched mapping for HIP/MUSA vendor headers via pull request #25033.
• Disabled KleidiAI support for macOS Apple Silicon builds.
• Provided binaries for Ubuntu x64/arm64/s390x (CPU, Vulkan, ROCm 7.2, OpenVINO, SYCL FP32/FP16).
• Included Windows builds for CPU, OpenCL Adreno, CUDA 12.4/13.3, Vulkan, OpenVINO, SYCL, and HIP.
• Released iOS XCFramework, Android arm64 (CPU), and macOS Apple Silicon/Intel binaries.

The release ensures broad hardware compatibility by offering specific builds for various accelerators like ROCm, SYCL, and OpenVINO alongside standard CPU versions.