[fix](test) Set profile level for writer profile case

[shuke987]

Summary

Fix branch-4.1 query_profile/test_writer_profile by explicitly setting profile_level=1 before the first INSERT profile check.

The case verifies that writer details are hidden at profile level 1 and present at profile level 2. branch-4.1 now defaults profile_level to 2, so the first half must set the intended level explicitly instead of depending on the old default.

Testing

• git diff --check
• Runtime regression test not run locally; requires branch-4.1 regression environment with S3 credentials

[shuke987]

run buildall

[shuke987]

review

[shuke987]

/review

[github-actions]

Review opinion: no blocking issues found.

I reviewed the PR diff and the relevant profile/session propagation paths. The change is limited to regression-test/suites/query_profile/test_writer_profile.groovy, adding set profile_level=1 before the negative writer-profile assertion and restoring a trailing newline. This matches the backend behavior: writer detail profiles are emitted only when enable_profile is true and profile_level >= 2, while profile_level defaults to 2 and is forwarded through SessionVariable.toThrift().

Critical checkpoints:

• Changed-file scope: GitHub reports only regression-test/suites/query_profile/test_writer_profile.groovy changed.
• Existing review context: no existing inline review comments or replies were present in the prepared context, and the PR comments API returned none before submission.
• AGENTS/code-review skill: no required AGENTS.md files were listed. I searched for a repo-local SKILL.md/code-review skill in this checkout before diff review, but none was present, so the prompt rules were used directly.
• Correctness/session behavior: the added session setting is consistent with the intended test isolation for the first profile check; the second half explicitly sets profile_level=2 before checking detailed writer profile output.
• Duplicate/unresolved candidates: none remained after the final sweep.
• Validation: static review only. The local checkout is grafted/shallow (HEAD^ is unavailable), so I used GitHub PR metadata/patch as the authoritative diff source. I did not run the S3-backed regression suite in this runner.

User focus: no additional user-provided review focus was supplied.

Subagent conclusions:

• optimizer-rewrite: no optimizer/rewrite correctness candidate; initial pass and convergence both returned NO_NEW_VALUABLE_FINDINGS.
• tests-session-config: no test/session/config candidate; initial pass and convergence both returned NO_NEW_VALUABLE_FINDINGS.
• Convergence round 1 ended with both live subagents returning NO_NEW_VALUABLE_FINDINGS for the same empty proposed final comment set.