llama.cpp b9936 release respects min-step when splitting prompt batches

The llama.cpp project has released version b9936, which includes a fix to respect the min-step parameter when splitting prompt batches.

• macOS Apple Silicon and Intel binaries are available, with KleidiAI support disabled on Apple Silicon.
• Linux builds cover Ubuntu x64, arm64, and s390x architectures for CPU, Vulkan, ROCm 7.2, OpenVINO, and SYCL backends.
• Android arm64 (CPU) binaries are provided.
• Windows builds support x64 and arm64 CPUs, OpenCL Adreno, CUDA 12/13, Vulkan, OpenVINO, SYCL, and HIP.
• openEuler support is disabled for the main branch but available for specific hardware configurations like 310p and 910b ACL Graph.
• A standalone UI binary is also included in the release.