refactor EAGLE 2 (#3269)

Co-authored-by: Ying Sheng <sqy1415@gmail.com>
Co-authored-by: merrymercy <lianminzheng@gmail.com>
Co-authored-by: Ying1123 <sqy1415@gmail.com>

examples/runtime/engine/EAGLE_offline_batch_inference.py

@@ -21,6 +21,7 @@ def main():
         speculative_num_steps=3,
         speculative_eagle_topk=4,
         speculative_num_draft_tokens=16,
+        cuda_graph_max_bs=8,
     )
 
     outputs = llm.generate(prompts, sampling_params)