llama.cpp

ggml : add ggml_backend_op_alloc_size_may_expand, use it in RPC

#27960

Merged