feat: add frequency penalty to inference sampling pipeline

- Add FrequencyPenaltyStrategy (logit -= penalty * count)
- Per-task rep_window for penalty history lookup
- Wire through engine, task, executor, API layer
- Add --frequency_penalty and --rep_window to stream_chat.py
- 9 unit tests for frequency penalty strategy

astrai/inference/api/openai.py

@@ -21,7 +21,6 @@ logger = logging.getLogger(__name__)
 _UNSUPPORTED_PARAMS = (
     "n",
     "presence_penalty",
-    "frequency_penalty",
     "logit_bias",
     "user",
 )

astrai/inference/api/protocol.py

@@ -125,6 +125,7 @@ class ProtocolHandler:
             temperature=self.request.temperature,
             top_p=self.request.top_p,
             top_k=self.request.top_k,
+            frequency_penalty=getattr(self.request, "frequency_penalty", 0.0),
         )
 
         if self.request.stream: