[Perf] Enhance benchmark script to support baseline thresholds and proved result handling (#2789)

tests/dfx/stability/scripts/test_benchmark_stability.py

@@ -112,6 +112,8 @@ def _run_one_benchmark_batch(
             flow=flow,
             dataset_name=dataset_name,
             num_prompt=num_prompts,
+            random_input_len=params.get("random_input_len"),
+            random_output_len=params.get("random_output_len"),
         )
         return result
     except (FileNotFoundError, OSError) as e: