llama.cpp b9548 patch fixes vocab compatibility across platforms

The open-source llama.cpp project, widely used for running large language models locally, has shipped version b9548. The release primarily addresses a vocabulary compatibility check that could cause mismatches when loading certain models. This fix, referenced in pull request #24256, ensures that tokenizer vocabularies are correctly validated before inference begins. The patch is available for all major platforms: macOS (both Apple Silicon and Intel x64, plus iOS XCFramework), Linux (x64, arm64, s390x CPU; Vulkan; ROCm 7.2; OpenVINO; SYCL FP32), Android (arm64 CPU), and Windows (x64 and arm64 CPU, CUDA 12 and 13, Vulkan, HIP). The project, which has over 115,000 stars on GitHub, continues to support bleeding-edge hardware acceleration like KleidiAI on Apple Silicon and AMD ROCm.

In addition to the core fix, the release includes pre-built binaries for all supported configurations, making it easy for developers and power users to update. Notably, some builds (like Apple Silicon with KleidiAI, SYCL FP32, and openEuler variants) are marked as DISABLED in this release, indicating ongoing work. The UI assets have also been updated. For professionals running LLMs locally—whether for prototyping, privacy, or latency—this patch is a quality-of-life improvement that reduces cryptic errors during model loading. The GitHub release was signed with a verified GPG key, ensuring integrity.