[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/conftest.py

@@ -280,7 +280,7 @@ class VllmRunner:
         disable_log_stats: bool = True,
         tensor_parallel_size: int = 1,
         block_size: int = 16,
-        enable_chunked_prefill: bool = False,
+        enable_chunked_prefill: bool = True,
         swap_space: int = 4,
         enforce_eager: Optional[bool] = False,
         quantization: Optional[str] = None,