llama.cpp b10015: OpenCL fix and expanded platform support

The llama.cpp project, a widely-used C++ implementation for running large language models locally, has released version b10015. This release is primarily a maintenance update that addresses a specific OpenCL compatibility issue. The commit note states: "opencl: do not use clCreateBufferWithProperties when targeting CL 2.x (#25673)". This fix ensures that users running older OpenCL 2.x drivers (common on some GPUs and integrated graphics) can properly allocate GPU memory without errors, improving stability for local LLM inference.

The new version also comes with an extensive set of pre-built binaries across 27 different targets, covering macOS (Apple Silicon with and without KleidiAI, Intel x64), multiple Linux architectures (x64, ARM, s390x) with various backends (Vulkan, ROCm 7.2, OpenVINO, SYCL FP32/FP16), Windows (x64 CPU, ARM64 CPU, ARM64 with OpenCL Adreno, CUDA 12 and 13, Vulkan, HIP), Android ARM64, iOS XCFramework, and openEuler with ACL Graph. This breadth ensures that developers and end-users can run llama.cpp on nearly any modern hardware configuration, whether for development, research, or production deployment of local AI models.