llama.cpp

fix tokenizer for JetBrain Mellum

#15045

Merged