Fix the value issue of anthropic (#3351)

### What problem does this PR solve?

This pull request fixes the issue mentioned in
https://github.com/infiniflow/ragflow/issues/3263.

1. response should be parsed as dict, prevent the following code from
failing to take values:
ans = response["content"][0]["text"]
2. API Model ```claude-instant-1.2``` has retired (by
[model-deprecations](https://docs.anthropic.com/en/docs/resources/model-deprecations)),
it will trigger errors in the code, so I deleted it from the
conf/llm_factories.json file and updated the latest API Model
```claude-3-5-sonnet-20241022```



### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

---------

Co-authored-by: chenhaodong <chenhaodong@ctrlvideo.com>
Co-authored-by: Kevin Hu <kevinhu.sh@gmail.com>

conf/llm_factories.json

@@ -2371,8 +2371,8 @@
                     "model_type": "chat"
                 },
                 {
-                    "llm_name": "claude-instant-1.2",
-                    "tags": "LLM,CHAT,100k",
+                    "llm_name": "claude-3-5-sonnet-20241022",
+                    "tags": "LLM,CHAT,200k",
                     "max_tokens": 102400,
                     "model_type": "chat"
                 }

rag/llm/chat_model.py

@@ -1260,7 +1260,7 @@ class AnthropicChat(Base):
                 system=self.system,
                 stream=False,
                 **gen_conf,
-            ).json()
+            ).to_dict()
             ans = response["content"][0]["text"]
             if response["stop_reason"] == "max_tokens":
                 ans += (