Fix top_p: null still sent in API requests after #385 (#415)

PR #385 made top_p optional and defaulted it to None in LLMConfig,
but openai.py still unconditionally includes top_p in the request
params dict. This sends "top_p": null to providers like AWS Bedrock
that don't strip null values, causing 400 errors.

Only include top_p in params when it is explicitly set.

Fixes #414

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Author: toslali-ibm

openevolve/llm/openai.py

@@ -155,9 +155,11 @@ async def generate_with_context(
                 "model": self.model,
                 "messages": formatted_messages,
                 "temperature": kwargs.get("temperature", self.temperature),
-                "top_p": kwargs.get("top_p", self.top_p),
                 "max_tokens": kwargs.get("max_tokens", self.max_tokens),
             }
+            top_p = kwargs.get("top_p", self.top_p)
+            if top_p is not None:
+                params["top_p"] = top_p
 
             # Handle reasoning_effort for open source reasoning models.
             reasoning_effort = kwargs.get("reasoning_effort", self.reasoning_effort)