llama.cpp b9995 sets SYCL threads to 256 for Battlemage and disables KleidiAI

The llama.cpp project released version b9995, which includes a specific optimization for Intel's Battlemage architecture and changes to macOS support.

• The SYCL backend now sets the `fattn_vec_nthreads` parameter to 256 for Battlemage (Xe2) and Lunar Lake GPUs, while keeping the default at 128 for Intel ARC Alchemist.
• KleidiAI support on macOS Apple Silicon has been disabled in this release.
• Binaries are provided for macOS (Apple Silicon and Intel), Linux (CPU, Vulkan, ROCm 7.2, OpenVINO, SYCL FP32/FP16), Android, Windows (CPU, OpenCL Adreno, CUDA 12/13, Vulkan, OpenVINO, SYCL, HIP), and openEuler.

This update ensures proper thread configuration for newer Intel hardware architectures while adjusting platform-specific feature availability.