llama.cpp b9886 adds UE4M3 LUT to ARM NVFP4 dot product

The llama.cpp project released version b9886, introducing a technical update to the ggml-cpu backend. The primary change is the use of a UE4M3 Look-Up Table (LUT) in the ARM NVFP4 dot product operation.

This release provides binaries for macOS Apple Silicon and Intel, Linux Ubuntu across CPU, Vulkan, ROCm, OpenVINO, and SYCL backends, as well as Windows with CUDA 12/13, Vulkan, OpenCL, and HIP support. Android arm64 builds are also available alongside the standard UI.

KleidiAI support for macOS Apple Silicon has been disabled in this build.