llama.cpp

Fix more int overflow during quant (PPL/CUDA).

#6563

Merged