llama.cpp b9594 adds accent stripping and vocab normalization refactor

llama.cpp, the popular open-source C++ library for running large language models locally, has released version b9594. The marquee change is a refactor of the vocabulary normalizer flags into a dedicated options struct, which now includes a new `strip_accents` feature. This allows developers to optionally strip diacritical marks from input text during tokenization, critical for languages like French, Spanish, or Vietnamese where accented characters can inflate token counts or cause mismatches.

The release also demonstrates the project's broad platform coverage: pre-built binaries are available for macOS (Apple Silicon and Intel, plus iOS XCFramework), Linux (x64 and arm64 with CPU, Vulkan, ROCm 7.2, and OpenVINO backends), Windows (x64 and arm64 with CPU, CUDA 12/13, Vulkan, and HIP), Android (arm64), and even openEuler with ACL Graph. The commit is signed with GitHub's verified signature. Contributors Sigbjørn Skjæret and the llama.cpp team collaborated on the changes, which are now merged into the main branch.