Order JDBC principal session rows by primary key

[AlexanderReaper7]

The default principal-name query can return a session's attribute rows in separate groups. SessionResultSetExtractor combines only adjacent rows, so a later group creates a partial session that replaces the earlier one in the result map.

Add ORDER BY S.PRIMARY_ID to keep every session's rows adjacent. A PostgreSQL integration test saves 120 sessions, adds an attribute to every third session in a later save, runs ANALYZE, and checks every listed session's attribute names and values against findById. Without the ordering, the regression fails because a listed session contains only later instead of all five stored attributes. With the ordering, it passes.

Closes #3452

Targets 3.5.x, the oldest supported maintenance branch listed in the open milestones, as required by CONTRIBUTING.adoc.

Validation ran in Docker with Temurin 17.0.20.1 and the repository's PostgreSQL 14.5 container:

• The new regression fails without the fix and passes with it.
• All 142 JDBC unit tests and 38 PostgreSQL repository integration tests pass.
• checkstyleMain, checkstyleIntegrationTest and checkFormat pass.

Other database integration suites and the full multi-module build were not run.