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/tests/integration_tests/llms/test_rwkv.py

@@ -31,5 +31,5 @@ def test_rwkv_inference() -> None:
     """Test valid gpt4all inference."""
     model_path = _download_model()
     llm = RWKV(model=model_path, tokens_path="20B_tokenizer.json", strategy="cpu fp32")
-    output = llm("Say foo:")
+    output = llm.invoke("Say foo:")
     assert isinstance(output, str)