llama : second attempt to refactor vision API

src/llama-model.h

@@ -4,6 +4,7 @@
 #include "llama-arch.h"
 #include "llama-hparams.h"
 #include "llama-vocab.h"
+#include "llama-vision.h"
 
 #include <memory>
 #include <string>
@@ -362,6 +363,10 @@ struct llama_model {
 
     const struct ggml_tensor * get_tensor(const char * name) const;
 
+    // vision
+    bool has_vision = false;
+    clip_vision_model clip;
+
 private:
     struct impl;
     std::unique_ptr<impl> pimpl;