Fixes #596 - Hardcoded model deployment name in azure_openai_complete

Fixes #596

Update `azure_openai_complete` function to accept a model parameter with a default value of 'gpt-4o-mini'.

* Modify the function signature of `azure_openai_complete` to include a `model` parameter with a default value of 'gpt-4o-mini'.
* Pass the `model` parameter to the `azure_openai_complete_if_cache` function instead of the hardcoded model name 'conversation-4o-mini'.

---

For more details, open the [Copilot Workspace session](https://copilot-workspace.githubnext.com/HKUDS/LightRAG/issues/596?shareId=XXXX-XXXX-XXXX-XXXX).

lightrag/llm.py

@@ -622,11 +622,11 @@ async def nvidia_openai_complete(
 
 
 async def azure_openai_complete(
-    prompt, system_prompt=None, history_messages=[], keyword_extraction=False, **kwargs
+    model: str = "gpt-4o-mini", prompt, system_prompt=None, history_messages=[], keyword_extraction=False, **kwargs
 ) -> str:
     keyword_extraction = kwargs.pop("keyword_extraction", None)
     result = await azure_openai_complete_if_cache(
-        "conversation-4o-mini",
+        model,
         prompt,
         system_prompt=system_prompt,
         history_messages=history_messages,