llama.cpp adds Ling 3.0 parser support

The llama.cpp project has released version b11057, adding a specialized chat parser for the Ling 3.0 (Bailing V3) model. This update addresses a critical parsing failure where the model's unique 'pre-opened' thinking block caused tool calls to be incorrectly interpreted as reasoning content, effectively breaking autonomous agent workflows. The new parser aligns llama.cpp’s behavior with reference implementations in vLLM and SGLang, ensuring accurate handling of mixed reasoning and tool-use turns. This release also includes standard binary updates for CUDA 13, ROCm 10.0, and various Apple Silicon configurations.