llama.cpp

[Model] Qwen3.5 dense and MoE support (no vision)

#19435

Merged