further shaving int4 layers to improve e2e test time

Summary

Signed-off-by: HDCharles <charlesdavidhernandez@gmail.com>

Author: HDCharles

tests/e2e/vLLM/configs/qwen3_w4a16_grouped_quant.yaml

@@ -6,4 +6,4 @@ dataset_id: HuggingFaceH4/ultrachat_200k
 dataset_split: train_sft
 num_calibration_samples: 20
 
-recipe: tests/e2e/vLLM/recipes/WNA16/recipe_w4a16_group_quant_first_20_layers.yaml
+recipe: tests/e2e/vLLM/recipes/WNA16/recipe_w4a16_group_quant_first_10_layers.yaml