Stable 25.8 Backport of #85455 - Propagate query plan optimizations to correlated subquery input

[mkmkme]

Fixes #2051

Changelog category (leave one):

• Improvement

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

Make query plan optimizations visible to the correlated subquery input subplan by postponing it's materialization (ClickHouse#85455 by @novikd).

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 [cf43375]

[mkmkme]

the fast test failed due to changes in this PR. Investigating

[CarlosFelipeOR]

The Stateless tests (amd_binary, ParallelReplicas, s3 storage, parallel) job crashes the server (server died) on this PR, on both the initial run and the re-run.

In both runs the server aborts inside ParallelReplicasReadingCoordinator (src/Storages/MergeTree/ParallelReplicasReadingCoordinator.cpp), and in both the crashing query is the same: SELECT count() FROM t_constraint_corr_root (from 04299_constraint_subst_correlated_subquery_root.sql, a correlated-subquery test).

• Run 1 (job): Logical error: 'Initiator received more initial requests than there are replicas: replica_num=0'. Reported failing test: 02149_schema_inference_formats_with_schema_3.
• Run 2 (re-run) (job): Logical error: 'Duplicate announcement received for replica number 0'. Reported failing test: 02751_multiquery_with_argument.

The reported failing test differs between the two runs (02149… vs 02751…) because it is simply whichever test was running in parallel when the server died — not the cause.

This server crash on ParallelReplicas is not present on the base branch stable-25.8 nor on any other PR in the CI database (gh-data.checks) over the last 90 days — it appears only on this PR.

@mkmkme we need to investigate and resolve this before merging.

[mkmkme]

Ilia raised a (very valid) concern that this PR introduces too many changes at the same time for no good reason. I totally agree with him.

From my point of view, I wanted to publish it to see whether it would break anything. Apparently it does break quite a few things, and I'm getting less and less confident in its state.

Let's close it.