[core][flink] Support view operations in JDBC Catalog

[tchivs]

Revives #6667 and ports the JDBC catalog view support onto current master.

Hi @JingsongLi, I addressed the questions from the previous review in #6667:

• Added automatic initialization of the JDBC view metadata table paimon_views.
  It stores (catalog_key, database_name, view_name, view_schema) with a primary key on the catalog/database/view tuple.
• Stored views as serialized Paimon ViewSchema; the JDBC catalog does not parse or resolve the SQL text during createView.
• Added tests for views referencing nonexistent tables, including a nonexistent table in another database.
• Added coverage for cross-database view rename.
• Added concurrent createView coverage relying on the JDBC primary key constraint.
• Updated the implementation for the current ViewChange API and listViewDetailsPaged fallback.
• Kept the SQLite dependency test-scoped for the Flink integration test.

Validation:

• git diff --check
• mvn -pl paimon-core spotless:check
• mvn -pl paimon-core -am -Pfast-build -DfailIfNoTests=false -Dtest=JdbcCatalogTest test
  • JdbcCatalogTest ran 62 tests, 0 failures, 0 errors.

I also attempted:

• mvn -pl paimon-flink/paimon-flink-common -am -Pfast-build -DfailIfNoTests=false -Dtest=JdbcCatalogViewITCase test

That command was blocked before test execution by dependency resolution timeout from https://jindodata-binary.oss-cn-shanghai.aliyuncs.com/mvn-repo/ for com.aliyun.jindodata:jindo-core:6.9.1 and jindo-sdk:6.9.1 while building unrelated paimon-jindo.

[tchivs]

@JingsongLi I pushed a follow-up commit to address the old review questions from #6667 and a few local review edge cases. PTAL when you have time.

What changed:

• The JDBC catalog initializes a paimon_views metadata table automatically. It stores (catalog_key, database_name, view_name, view_schema) with a primary key on the catalog/database/view tuple.
• View definitions are stored as serialized Paimon ViewSchema; JDBC catalog does not parse or resolve the SQL text during createView. This is covered by tests for nonexistent table references and cross-database references.
• Cross-database renameView is supported and tested.
• Concurrent duplicate view creation is covered by a two-thread test and relies on the JDBC metadata primary key. I also centralized uniqueness violation detection for SQLite message text and common SQLState-based drivers.
• Added hardening for catalog invariants: non-cascade dropDatabase now rejects view-only databases, renameView rejects missing target databases, and table/view same-name collisions are rejected on create and rename paths.
• Added JDBC catalog view documentation in concepts, Flink SQL DDL, and Spark SQL DDL docs.

Local validation:

• git diff --check
• mvn -pl paimon-core spotless:check
• mvn -pl paimon-core -am -Pfast-build -DfailIfNoTests=false -Dtest=JdbcCatalogTest test -- 67 tests, 0 failures
• mvn -pl paimon-core -am -DfailIfNoTests=false -Dtest=JdbcCatalogTest test -- 67 tests, 0 failures

I also retried mvn -pl paimon-flink/paimon-flink-common -am -Pfast-build -DfailIfNoTests=false -Dtest=JdbcCatalogViewITCase test, but it was blocked before test execution by the external Jindo dependency repository timeout while resolving com.aliyun.jindodata:jindo-core:6.9.1 / jindo-sdk:6.9.1 for unrelated paimon-jindo.