llama.cpp adds support for vLLM d3 models

The llama.cpp repository has added support for vLLM d3 models, expanding local inference capabilities. This enables more efficient local execution of models originally optimized for the vLLM runtime.