llama.cpp

fix: use int64_t to avoid overflows in int schema to grammar conversion

#16626

Merged