Fix: Always pass reflection_lm_callable to GEPA - was broken when LLEGO disabled

The bug: When use_llego_operators=False, reflection_lm was set to None,
which meant GEPA could not generate any new candidate prompts.

Fix: Always pass the reflection_lm_callable to GEPA regardless of LLEGO mode.
This allows GEPA reflection to work even without LLEGO genetic operators.

src/gepa_optimizer/core/optimizer.py

@@ -896,10 +896,10 @@ Output the improved prompt directly and only the prompt."""
             
             # Prepare optimization parameters with ONLY valid GEPA parameters
             # Note: 'adapter' variable is set above (either LLEGO-enhanced or standard)
-            # 🔥 REMOVED: Excessive diagnostic warnings - moved to DEBUG level
-            reflection_lm_passed = reflection_lm_callable if self.config.use_llego_operators else None
-            if reflection_lm_passed:
-                self.logger.debug(f"reflection_lm_callable passed to GEPA (may be ignored in adapter mode)")
+            # 🔥 FIX: ALWAYS pass reflection_lm_callable to GEPA (required for generating new candidates!)
+            # Previously this was only passed when use_llego_operators=True, which broke standard reflection
+            reflection_lm_passed = reflection_lm_callable  # Always pass reflection callable
+            self.logger.info(f"✅ reflection_lm_callable passed to GEPA (LLEGO={self.config.use_llego_operators})")
             
             # Debug logging removed - not needed in production