update default gpt-3.5-turbo models

by default, only OpenAI was enabled, more models enable for more flexibility.

new provider `Koala` added, to watch out as it could be unstable.

g4f/client/client.py

@@ -96,6 +96,7 @@ class Completions():
             ignore_working,
             ignore_stream,
         )
+        
         stop = [stop] if isinstance(stop, str) else stop
         response = provider.create_completion(
             model, messages,

g4f/models.py

@@ -19,6 +19,7 @@ from .Provider   import (
     Llama2,
     Vercel,
     Gemini,
+    Koala,
     Bing,
     You,
     Pi,
@@ -71,7 +72,12 @@ gpt_35_long = Model(
 gpt_35_turbo = Model(
     name          = 'gpt-3.5-turbo',
     base_provider = 'openai',
-    best_provider = OpenaiChat
+    best_provider = RetryProvider([
+        FreeGpt,
+        You,
+        ChatgptNext,
+        Koala,
+    ])
 )
 
 gpt_4 = Model(