[CI]enable chunked prefill by default (#4569)

set `enable_chunked_prefill` to True for e2e test by default to keep the
same behavior with vLLM

- vLLM version: v0.11.2

Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com>

tests/e2e/singlecard/spec_decode_v1/test_v1_spec_decode.py

@@ -118,7 +118,6 @@ def test_eagle_correctness(
     spec_model_name = eagle3_model_name() if use_eagle3 else eagle_model_name()
     with VllmRunner(
             model_name,
-            enable_chunked_prefill=True,
             max_num_seqs=1,
             max_num_batched_tokens=2048,
             gpu_memory_utilization=0.6,