patch: remove usage of llm, chat model __call__ (#20788)

- `llm(prompt)` -> `llm.invoke(prompt)`
- `llm(prompt=prompt` -> `llm.invoke(prompt)` (same with `messages=`)
- `llm(prompt, callbacks=callbacks)` -> `llm.invoke(prompt,
config={"callbacks": callbacks})`
- `llm(prompt, **kwargs)` -> `llm.invoke(prompt, **kwargs)`

cookbook/Multi_modal_RAG_google.ipynb

@@ -256,7 +256,7 @@
     "    \"\"\"Make image summary\"\"\"\n",
     "    model = ChatVertexAI(model_name=\"gemini-pro-vision\", max_output_tokens=1024)\n",
     "\n",
-    "    msg = model(\n",
+    "    msg = model.invoke(\n",
     "        [\n",
     "            HumanMessage(\n",
     "                content=[\n",