Potential bug in`llm_judge_column`

Inside function `llm_judge_column`, it seems that the following code will not be executed, since `len(score_list) == len(response)` in current implementation.
```
if len(score_list) != len(response):
score_list = [0] * len(response)
msg_list = ["llm judge failed due to length"] * len(response)
```
https://github.com/ByteDance-Seed/WideSearch/blob/9825ba7b140b71d81b364793f86dabe4cfed6749/src/evaluation/metric_utils.py#L376