feat: using the gpt-4o-mini instead of gpt3.5-turbo!

it's cheaper and have a better performance.

subquery.py

@@ -200,7 +200,7 @@ def query_multiple_source(
         )
 
     embed_model = CohereEmbedding()
-    llm = OpenAI("gpt-3.5-turbo")
+    llm = OpenAI("gpt-4o-mini")
     Settings.embed_model = embed_model
     Settings.llm = llm