Fix relational JDBC location-overlap index to match the optimized sibling-check query

[zhang-arvin]

The optimized sibling check reads location_without_scheme via QueryGenerator.generateOverlapQuery, which filters realm_id and catalog_id. On Postgres and CockroachDB the idx_locations index led with parent_id instead of catalog_id, so the predicate could only probe the realm_id prefix and every CREATE TABLE / CREATE NAMESPACE with OPTIMIZED_SIBLING_CHECK enabled fell back to a scan instead of the intended indexed lookup. H2 already used the correct columns.

Schema v5 has already shipped in 1.7.0, so rather than editing a released script this adds schema v6 for all backends with the corrected index and bumps the latest schema version. Fresh installs bootstrap at v6 and get the right index automatically. Existing deployments need a one-time manual index recreation, documented in the changelog upgrade notes, because Polaris has no automated schema migrations.

Changes

• DatabaseType.java: Bump latest schema version from 5 to 6 for all backends
• schema-v6.sql (postgres/cockroachdb): Fix idx_locations from (realm_id, parent_id, location_without_scheme) to (realm_id, catalog_id, location_without_scheme)
• schema-v6.sql (h2): No-op version bump to keep H2 in sync (already correct)
• CHANGELOG.md: Document the fix and upgrade notes for existing deployments
• SchemaLocationIndexParityTest.java: New test to ensure idx_locations column list stays consistent across backends

Fixes #5301

[vigneshio]

Thanks @zhang-arvin. This is the same change as #5301, which is already under review. please close this as a duplicate.

if you'd like to help, reviewing #5301 would be the right place..

[zhang-arvin]

Closing as duplicate of #5301. Thanks for the review!