llama.cpp b11045 adds CUDA 13 and ROCm 10.0 builds

llama.cpp has released version b11045, expanding its pre-built binary support to include CUDA 13 (versions 12.8 on Linux, 13.3/13.4 on Windows) and ROCm 10.0 for both AMD GPUs and openEuler systems. The update also adds OpenVINO and SYCL backends for Intel hardware, while disabling macOS KleidiAI and openEuler default builds in this iteration. These additions allow developers to run local LLMs on newer NVIDIA and AMD hardware without manual compilation. The release reinforces llama.cpp's position as the standard inference engine for cross-platform AI deployment.