llama.cpp b9957 improves server tools and removes apply_diff

The llama.cpp project released version b9957, which focuses on improving the server's tooling infrastructure. The update introduces a `tools_io` abstraction and its basic implementation while removing the `apply_diff` function.

• Improved edit tool functionality
• Added tools_io abstraction and tools_io_basic
• Moved utils to class member and added const qualifiers
• Fixed build issues
• Disabled macOS Apple Silicon KleidiAI support
• Provided binaries for macOS, Linux, Android, Windows, and openEuler across CPU, GPU, and various accelerators like CUDA, Vulkan, ROCm, and OpenVINO

This release updates the server-side capabilities and provides updated binaries for a wide range of hardware platforms.