fix: show error details in verbose benchmark output

When a sample errors during benchmark execution, the verbose output
now shows the actual error message, making debugging easier:

  [✗] Sample niah-0000: {'correct': 0.0, 'f1': 0.0}
      Error: <actual error message here>

Author: claude

benchmarks/runner.py

@@ -292,6 +292,8 @@ def _update_progress(
         if self.config.verbose:
             status = "✓" if sample_result.is_correct else "✗"
             print(f"  [{status}] Sample {sample_result.sample_id}: {sample_result.metrics}")
+            if sample_result.error:
+                print(f"      Error: {sample_result.error}")
 
     def _run_sequential(
         self,