llama.cpp

docker: parameterize CUDA version in build matrix, remove cuda-new.Dockerfile

#20920

Merged