llama.cpp b10031 fixes tokenize --stdin mutual-exclusion bug

The open-source LLM inference engine llama.cpp (121k stars on GitHub) has released version b10031, a maintenance update addressing a specific but impactful CLI bug. The fix, merged in pull request #25672, drops the `--stdin` mutual-exclusion check from the `tokenize` command. Previously, attempting to pipe input to `tokenize` would trigger an error stating that `--stdin` was incompatible, even though the command should support it. This behavior was inconsistent with the `main` CLI and `completion` commands, which allow `--stdin`. The change brings `tokenize` in line with user expectations and the rest of the interface, reducing friction for developers and power users who rely on piped workflows.

This release is particularly relevant for users running large language models locally via llama.cpp, a popular tool for efficient inference on consumer hardware. By fixing this minor but annoying issue, the project continues to polish its command-line experience. The b10031 release includes pre-built binaries for a wide range of platforms: macOS (Apple Silicon, Intel, iOS), Windows (CPU, arm64, CUDA 12/13, Vulkan, OpenVINO, SYCL), Linux (CPU, arm64, Vulkan, ROCm 7.2, OpenVINO, SYCL), and Android arm64. As always, the source is available for custom builds. This update underscores llama.cpp's commitment to incremental improvements that enhance usability for the local AI community.