llama.cpp Implements hparams.n_layer_nextn Loading Fix

The llama.cpp repository merged a fix to load hparams.n_layer_nextn before n_layer calls. This ensures correct model architecture parameter initialization during the loading process.