Fix swapped names in PostgreSQL undefined column error message

[thswlsqls]

Fixes #39123.

Changes proposed in this pull request:

• Swap the two arguments passed to PostgreSQLVendorError.UNDEFINED_COLUMN in PostgreSQLDialectExceptionMapper.convert() so the reason format Column "%s" of table "%s" does not exist receives the column name first and the table name second. A missing column user_id in table t_order now produces Column "user_id" of table "t_order" does not exist instead of the reversed message.
• Aligns this branch with sibling branches in the same mapper (PRIVILEGE_NOT_GRANTED, INVALID_PARAMETER_VALUE, PROTOCOL_VIOLATION) that already match field order to format order.
• Add assertConvertColumnNotFoundMessage asserting the formatted message body; the existing assertConvert only checked sqlState and severity, so it missed the swap.

---

Before committing this PR, I'm sure that I have checked the following options:

• My code follows the code of conduct of this project.
• I have self-reviewed the commit code.
• I have (or in comment I request) added corresponding labels for the pull request.
• I have passed maven check locally : ./mvnw clean install -B -T1C -Dmaven.javadoc.skip -Dmaven.jacoco.skip -e.
• I have made corresponding changes to the documentation.
• I have added corresponding unit tests for my changes.
• I have updated the Release Notes of the current development version. For more details, see Update Release Note

[terrymanu]

Summary

Review Result: Not Mergeable

Feedback Mode: Needs Discussion

Reason: The patch correctly reorders arguments for the dialect exception in isolation, but that exception is not produced by the PostgreSQL SELECT binding path described in #39123. The linked issue’s root-cause model and the implementation scope need to be reconciled before this can close the reported bug.

Issues

P1: The changed mapper is not on the reported SELECT failure path

• Problem: #39123 reports selecting a nonexistent column. However, the relevant dialect ColumnNotFoundException is constructed by PostgreSQL only inside the prepared-statement INSERT Describe branch at PostgreSQLComDescribeExecutor.java:106. A normal projected column is bound through ColumnProjectionSegmentBinder.java:46, where ColumnSegmentBinder.java:216 throws the separate infra ColumnNotFoundException. That exception extends ShardingSphereSQLException and is converted before the dialect-mapper branch at SQLExceptionTransformEngine.java:49. The new test at PostgreSQLDialectExceptionMapperTest.java:83 mocks the mapper input directly, so it does not establish the reported Proxy path.
• Impact: Merging with Fixes #39123 could close the reported SELECT bug while only changing the extended-protocol INSERT Describe behavior.
• Discussion Needed: Please pause the current implementation direction and confirm the exact SQL and protocol phase that reproduces #39123. If it is a SELECT, the discussion should identify the binder/error-transformation owner. If it is actually an INSERT Describe failure, the issue and PR scope should first be narrowed accordingly. The type: discussion label is appropriate until this path is agreed.

Review Details

• Review Focus: Code Correctness Review — CI not reviewed by request.
• Reviewed Scope: All three changed files: RELEASE-NOTES.md, PostgreSQLDialectExceptionMapper.java, and PostgreSQLDialectExceptionMapperTest.java; supporting exception definitions, production construction sites, PostgreSQL Describe handling, column binding, and exception transformation were traced. Latest head: 0eddd07cef46ead1ee6bbab3f34107e13ee32221; base: master at 7568693e8ee02637a47d3cca8acbf449797932e5; local merge-base: 6778bfbe08c1ff59574874575a6ab4b932066dba. The local triple-dot file list matched GitHub’s /pulls/39125/files list.
• Not Reviewed Scope: GitHub Actions/CI and live Proxy/client reproduction were not reviewed. Unrelated modules were inspected only where necessary to trace the exception path.
• Verification: GitHub PR, file, commit, issue, comment, and review evidence was accessible. The bounded review inventory, merge-base comparison, exact-head source inspection, and repository-wide usage searches completed successfully. Maven was not run because the active checkout was not the PR head; no stale local test result was used. This does not affect the blocker because the path mismatch is established directly from the latest-head production call graph.
• Release Note / User Docs: A release-note entry is present, but its applicability depends on the resolved issue scope. Separate user documentation is not required because no configuration, API, migration, or compatibility contract changes.

[thswlsqls]

Agreed on the mechanics: the mapper only sees the dialect ColumnNotFoundException (database.exception.core…), which for PostgreSQL is constructed solely in PostgreSQLComDescribeExecutor's INSERT Describe branch. A projected SELECT column throws the infra ColumnNotFoundException via ColumnSegmentBinder, and SQLExceptionTransformEngine routes that through its ShardingSphereSQLException branch before the dialect-mapper branch, so it never hits this mapper.

That said, the swapped Column "x" of table "y" does not exist text is uniquely PostgreSQLVendorError.UNDEFINED_COLUMN from this mapper — the infra path emits Unknown column '…' in '…' instead. So #39123's symptom, if it shows that swapped wording, must traverse the extended-protocol Describe (INSERT) phase, and the patch fixes it there. I'll confirm the exact SQL/phase with the reporter and correct the "SELECT" wording. If a plain SELECT is also expected to surface this, that's a separate change in ColumnSegmentBinder, out of scope here.