llama.cpp

ggml: WebGPU backend host improvements and style fixing

#14978

Merged