model: using single llm_build per arch (#21970)

* model: using single llm_build per arch

* fix merge

* nits

src/models/t5encoder.cpp

@@ -0,0 +1,3 @@
+#include "models.h"
+
+llm_build_t5encoder::llm_build_t5encoder(const llama_model & model, const llm_graph_params & params) : llm_build_t5<true>(model, params) {}