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)`

libs/community/langchain_community/llms/weight_only_quantization.py

@@ -222,7 +222,7 @@ class WeightOnlyQuantPipeline(LLM):
                     model_id="google/flan-t5-large",
                     task="text2text-generation",
                 )
-                llm("This is a prompt.")
+                llm.invoke("This is a prompt.")
         """
         response = self.pipeline(prompt)
         if self.pipeline.task == "text-generation":