llama : move n_vocab from llama_sampler_data to penalty_sampler (#26520)

This matches how it is done for logit_bias and mirostat samplers, see
https://github.com/ggml-org/llama.cpp/pull/25262#discussion_r3703951151

src/llama-graph.cpp

@@ -3683,7 +3683,6 @@ void llm_graph_context::build_sampling() const {
             /*.probs       =*/ nullptr,
             /*.sampled     =*/ nullptr,
             /*.candidates  =*/ nullptr,
-            /*.n_vocab     =*/ logits_seq->ne[0],
         };
 
         assert(sampler->iface->backend_apply);