Closes: #1717

Problem. EXPORT PART/EXPORT PARTITION matched source and destination columns positionally, exactly like INSERT INTO dest SELECT * FROM src, and required the column count to match exactly. This is unnecessarily strict for a common
real scenario: a MergeTree table's schema grows over time (new trailing columns added), but older partitions still need to be exported to an external table (Iceberg/object storage) whose schema was fixed when it was created. Such exports were rejected outright with NUMBER_OF_COLUMNS_DOESNT_MATCH, with no way to just drop the newer columns and export the rest.

Solution. Added the export_merge_tree_part_schema_mismatch_mode setting (default strict, preserving the old behavior). The new ignore_extra_source_columns_by_position mode allows the source to have more columns than the destination: the extra trailing source columns (by declared/readable position, the same order INSERT SELECT uses) are dropped, both at synchronous validation time (verifyExportSchemaCastable, so a real mismatch still fails immediately at ALTER TABLE ... EXPORT time) and at actual data-export time (ExportPartTask::addExportConvertingActions, via a projection step before the existing positional CAST). The reverse direction — destination has more columns than source — is still always rejected in both modes.

Known follow-up. Ignored columns are still fully read and decompressed from disk before being discarded in the projection step, so no I/O is saved. Skipping the read up front isn't safe as a naive truncation: a kept
ALIAS/MATERIALIZED column can legally forward-reference an ignored one, so it would need a real dependency-graph walk instead. Left as a possible future optimization

Changelog category (leave one):

• Improvement

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

Added the export_merge_tree_part_schema_mismatch_mode setting. With ignore_extra_source_columns_by_position (default: strict), EXPORT PART/EXPORT PARTITION allows a source table with extra trailing columns — they are simply
ignored instead of failing with NUMBER_OF_COLUMNS_DOESNT_MATCH.

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)