[fix](regression) Disable auto compaction for physical row inspections

[shuke987]

What problem does this PR solve?

Issue Number: N/A

Related PR: #65211, #64508

Problem Summary:

Some unique-key MOW regression cases deliberately bypass the delete bitmap and assert historical physical rows, sequence values, skip bitmaps, and rowset versions. These assertions require the pre-compaction rowsets to remain available, but the affected tables leave automatic compaction enabled. Background compaction can therefore merge the expected old rowsets before qt_inspect or an equivalent raw-row query runs, while all user-visible SQL results remain correct.

This PR disables table-level automatic compaction only for the tables whose assertions intentionally depend on historical physical layout. It covers nine case files, including publish-conflict, read-from-old, delete-sign, sequence-column, and auto-increment physical-row checks. Golden results and user-visible correctness assertions are unchanged.

The source scan reviewed all master cases that set skip_delete_bitmap=true without already disabling automatic compaction. Four remaining files were intentionally excluded because the variable is used only for debug-mode validation, routine-load waiting, insert execution setup, or binlog TVF behavior; they do not assert historical table rowsets.

Release note

None

Check List (For Author)

• Test

  • Regression test
  • Unit Test
  • Manual test (add detailed scripts or steps below)
  • No need to test or manual test. Explain why:
    • This is a refactor/code format and no logic has been changed.
    • Previous test can cover this change.
    • No code files have been changed.
    • Other reason

  Validation: test_flexible_partial_update_publish_conflict_seq passed as one lock-serialized suite on the authorized branch-4.1 test cluster using the patched master case file. The run confirmed the new table property in SHOW CREATE TABLE and completed all logical and physical-row assertions with 0 failed suites. git diff --check origin/master...HEAD also passed.

• Behavior changed:

  • No.
  • Yes.

• Does this need documentation?

  • No.
  • Yes.

Check List (For Reviewer who merge this PR)

• Confirm the release note
• Confirm test cases
• Confirm document
• Add branch pick label

[shuke987]

/review

[shuke987]

run buildall

[hello-stephen]

Thank you for your contribution to Apache Doris.
Don't know what should be done next? See How to process your PR.

Please clearly describe your PR:

1. What problem was fixed (it's best to include specific error reporting information). How it was fixed.
2. Which behaviors were modified. What was the previous behavior, what is it now, why was it modified, and what possible impacts might there be.
3. What features were added. Why was this function added?
4. Which code was refactored and why was this part of the code refactored?
5. Which functions were optimized and what is the difference before and after the optimization?

[github-actions]

Reviewed the PR at f7f58a9a4c1ac49f1696ad7f6c25d5df0f94dfee against base 2c5f1574737970b6123c1f0ac24d14a36f366381. I do not see a blocking issue.

Checkpoint conclusions:

• Goal and proof: The PR's goal is to stabilize regression tests that intentionally inspect pre-compaction physical rows for unique-key MoW partial-update/publish behavior. The added table property is applied to tables that later use skip_delete_bitmap, skip_delete_sign, qt_inspect/inspect_rows, hidden version columns, or equivalent physical-row assertions.
• Scope: The patch is narrowly limited to regression table properties plus newline normalization. No production code or golden result rows are changed.
• Concurrency and lifecycle: No runtime concurrency code is changed. Existing publish/debug-point concurrency in the tests remains the same; the new property only prevents background compaction from removing rowsets that the tests intentionally inspect.
• Configuration and compatibility: disable_auto_compaction is an existing table property. The FE create-table path stores it in table metadata, shared-nothing tablets receive it through tablet schema creation, and cloud tablet meta/scheduler paths also honor it. The property remains applicable through stream-load publish and table truncate/reuse flows.
• Parallel paths and coverage: I checked related skip_delete_bitmap/qt_inspect/physical-row cases. The remaining no-property suites using skip_delete_bitmap are binlog TVF, insert execution setup, routine-load waiting, or MV debug-mode validation, not comparable historical-row inspection cases.
• Test result files: The corresponding .out files contain no DDL, SHOW CREATE TABLE, PROPERTIES, or disable_auto_compaction output, so the unchanged golden files are consistent with this patch.
• Data correctness, persistence, performance, and observability: No data-path, persistence-format, metric, or logging behavior is changed. The compaction disablement is scoped to regression test tables.

I did not run builds or regression tests in this review-only runner. The review was completed with one convergence round: the main review plus two full-review subagents and one risk-focused subagent all returned no new valuable findings, and there are no unresolved candidates or inline comments.

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