25.3.8 Backport of #88090 - Optimize ReplacingMergeTree is_deleted FINAL queries by adding a filter expression transform

[mkmkme]

Currently, all ranges in a query with FINAL clause on a ReplacingMergeTree table with is_deleted column have to be processed by FINAL merge transform.

Add a FilterTransform correspdoning to is_deleted = 0 for parts & ranges thus enabling parallel processing and skip FINAL Merge transform. The parts & ranges that qualify are 1) Partitions with only a single part when do_not_merge_across_partitions_select_final is True 2) Non-intersecting ranges identified by the existing split ranges into intersecting/non-intersecting optimization.

Performance example : ClickHouse#65180 (comment)

Changelog category (leave one):

• Performance Improvement

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

SELECT query with FINAL clause on a ReplacingMergeTree table with the is_deleted column now executes faster because of improved parallelization from 2 existing optimizations : 1) do_not_merge_across_partitions_select_final optimization for partitions of the table that have only a single part 2) Split other selected ranges of the table into intersecting / non-intersecting and only intersecting ranges have to pass through FINAL merging transform (ClickHouse#88090 by @shankar-iyer).

CI/CD Options

Exclude tests:

• Fast test
• Integration Tests
• Stateless tests
• Stateful tests
• Performance tests
• All with ASAN
• All with TSAN
• All with MSAN
• All with UBSAN
• All with Coverage
• All with Aarch64
• 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)
• Parquet (1.5h)
• RBAC (1.5h)
• SSL Server (1h)
• S3 (2h)
• Tiered Storage (2h)

[ianton-ru]

LGTM