llama.cpp

fix typo: `n_ctx_pre_seq` -> `n_ctx_per_seq`

#13221

Merged