Change Spark status to avoid getting FINISHED before getting FAILED

Now it is confusing when job status change FINISHED to FAILED.

Look for Spark job status handling in the codebase, likely in modules related to job scheduling or Spark integration. The issue mentions avoiding 'FINISHED' before 'FAILED', so examine state transition logic. Check for existing tests around job status to understand the current behavior and where changes are needed.

由索引模型根据 Issue 内容生成。