Antalya 25.8 Backport of #88090: Optimize ReplacingMergeTree is_deleted FINAL queries by adding a filter expression transform

[mkmkme]

Same as #1332 but for Antalya 25.8

Changelog category (leave one):

• Performance Improvement

Changelog entry (a user-readable short description of the changes that goes to CHANGELOG.md):

...

Documentation entry for user-facing changes

...

CI/CD Options

Exclude tests:

• Fast test
• Integration Tests
• Stateless tests
• Stateful tests
• Performance tests
• Aarch64 tests
• All with ASAN
• All with TSAN
• All with MSAN
• All with UBSAN
• All with Coverage
• All Regression
• Disable CI Cache

Regression jobs to run:

• Fast suites (mostly <1h)
• Aggregate Functions (2h)
• Alter (1.5h)
• Benchmark (30m)
• ClickHouse Keeper (1h)
• Iceberg (2h)
• LDAP (1h)
• OAuth (5m)
• Parquet (1.5h)
• RBAC (1.5h)
• SSL Server (1h)
• S3 (2h)
• S3 Export (2h)
• Swarms (30m)
• Tiered Storage (2h)

[github-actions]

Workflow [PR], commit [c4e3960]

[mkmkme]

The included test was passing even on clean antalya-25.8 which is concerning. Otherwise the PR and the backport were trivial and I don't see any other reason why it couldn't be merged. I'll self-verify once CI passes

[mkmkme]

@blau-ai

[blau-ai]

CI triage

Verdict: 5 red checks, 0 caused by this PR. The only failing tests are the Iceberg regression suites (iceberg_1, iceberg_2), which are pre-existing failures on the antalya-25.8 base branch — unrelated to a ReplacingMergeTree FINAL optimization. FinishCIReport is just the aggregate gate reflecting those reds. Everything else is green: all builds, Fast test, every Stateless/Integration/Stress shard, AST fuzzer, BuzzHouse, Compatibility, Grype scans, and all other regression suites (swarms, parquet, s3_export).

Failing checks

Check	Signature	Classification
RegressionTestsRelease / Iceberg (1) / iceberg_1	1 module errored — 486 scenarios (477 ok, 1 failed, 3 features errored)	pre-existing / infra
RegressionTestsRelease / Iceberg (2) / iceberg_2	1 module failed — 1583 scenarios (1577 ok, 3 failed, 6 features failed)	pre-existing / infra
Regression x86 iceberg_1 / iceberg_2	same runs, S3 report mirror	pre-existing / infra
FinishCIReport	exit code 1 (aggregate gate)	consequence of the Iceberg reds

Evidence it is not this PR

• Scope mismatch. The diff touches only src/Processors/QueryPlan/ReadFromMergeTree.cpp (the MergeTree FINAL read path) plus one stateless test 02490_replacing_merge_tree_is_deleted_column_transform_opt. Nothing in the Iceberg / DataLake code path. The new stateless test itself passed (Stateless amd_debug parallel/sequential all green).

• Same failure on every sibling backport into antalya-25.8, all touching unrelated code:

  • Antalya 25.8 Cherry-pick: Force mutation to recalculate indices when rewriting compact parts #2162 (mutation index recalc) — iceberg_1: 1 module errored, iceberg_2: 1 module failed
  • Antalya 25.8 Backport of #89367: Crash in IN function where columns have different types and many columns are involved #2164 (IN function crash) — same signature
  • Antalya 25.8 Backport of #92250 - Reduce INSERT/merges memory usage with wide parts for very wide tables by enabling adaptive write buffers #2166 (adaptive write buffers) — same signature
  • Antalya 25.8 Backport of #88090: Optimize ReplacingMergeTree is_deleted FINAL queries by adding a filter expression transform #2163 (this PR) — same signature

  Identical failure fingerprint across four unrelated PRs ⇒ the break lives on the base branch / regression environment, not in any of these changes.

Suggested next step

No PR-side fix is warranted — nothing here can affect Iceberg. Options:

1. Merge as-is once a reviewer signs off; the Iceberg suite is red for all current antalya-25.8 backports and should not block this one.
2. If a hard gate requires green, have the CI owner re-run Iceberg (1)/(2) or track the base-branch Iceberg breakage as its own issue.

I couldn't extract the exact failing scenario names — the detailed report lives on altinity-build-artifacts.s3.amazonaws.com (no network fetch available to me) and the Actions log stream errored on retrieval. But the cross-PR evidence above is decisive for the PR-caused vs. not question, which is what matters here.

_{🤖 automated CI triage by @blau-ai}

[ilejn]

The included test was passing even on clean antalya-25.8 which is concerning.

Same against official 25.8