tts : fix n_ubatch + make WavTokenizer cache-less (#13713)

ggml-ci

src/llama-model.cpp

@@ -13189,6 +13189,7 @@ llama_memory_i * llama_model::create_memory(const llama_memory_params & params,
         case LLM_ARCH_JINA_BERT_V2:
         case LLM_ARCH_NOMIC_BERT:
         case LLM_ARCH_NOMIC_BERT_MOE:
+        case LLM_ARCH_WAVTOKENIZER_DEC:
             {
                 res = nullptr;
             } break;