llama.cpp b9467 cleans up unused variable warnings across all platforms

The latest release of llama.cpp, tagged b9467, is a maintenance update aimed at silencing unused variable warnings across the entire codebase. This cleanup helps developers avoid noise during compilation and ensures the project maintains clean code standards. The release was signed with GitHub's verified signature and includes builds for a wide range of platforms, including macOS (Apple Silicon, Intel), Linux (CPU, Vulkan, ROCm, OpenVINO, SYCL), Windows (CPU, CUDA 12/13, Vulkan, HIP), Android (arm64 CPU), and openEuler (with ACL Graph support). Several configurations (e.g., KleidiAI on macOS, SYCL on Windows, and openEuler) are marked as DISABLED in this release, likely due to ongoing testing or build issues.

The primary change—removing unused variable warnings—does not introduce new functionality or performance improvements. However, it reflects the project's ongoing effort to maintain a high-quality, warning-free codebase, which is critical for the many developers who build llama.cpp from source for their own LLM inference workloads. The release includes UI assets and a single commit (1fd5f48) that triggered the update. For users, this is a straightforward upgrade that should compile without new warnings, making it easier to integrate into custom pipelines.