llama: Wire up Qwen3.5/Qwen3.5MoE tensors for NVFP4 support#20506

CISC merged 2 commits intoggml-org:masterggml-org/llama.cpp:masterfrom

michaelw9999:nvfp4-qwen35-nvfp4-loader-fixmichaelw9999/llama.cpp:nvfp4-qwen35-nvfp4-loader-fixCopy head branch name to clipboard