llama.cpp b10416 fixes stale UI caching for web interface

ggml-org's llama.cpp, the wildly popular C++ inference engine for LLaMA models (124k stars, 21.7k forks), released build b10416 on August 13. The headline fix addresses a subtle but impactful web server issue: index.html was being served with a `max-age=31536000` cache header and marked immutable, just like the hashed static assets. However, unlike those versioned assets, index.html's filename stays constant across builds. This meant any browser or reverse proxy that cached the page would lock onto an outdated build, showing users an old UI even after a server upgrade. The change (#27006) switches index.html to ETag-based revalidation, so unchanged builds still return a 304 Not Modified, while new builds immediately serve the fresh interface.

Beyond the caching fix, this release continues llama.cpp's tradition of broad platform support. Binaries are available for macOS (Apple Silicon arm64, including a KleidiAI-enabled variant, plus Intel x64 and iOS XCFramework), Linux (Ubuntu with CPU, Vulkan, ROCm 7.14, OpenVINO, SYCL FP32/FP16, plus s390x and arm64), Windows (CPU, arm64, OpenCL Adreno, CUDA 12.4 and CUDA 13.3 DLLs, Vulkan, OpenVINO, SYCL, and ROCm 7.14), Android (arm64 CPU), and openEuler builds. The release also includes UI assets, reflecting the project's growing emphasis on a first-party web interface. For developers running llama.cpp in production, this resolves a frustrating class of bugs where the server was updated but the client UI appeared stuck. It's a small, focused fix that dramatically improves the reliability of llama.cpp's server mode.