Deduplicate fetch join results at QueryDSL level

[zio0911]

Summary

• Deduplicate results in fetch(), fetchOne(), and fetchResults() when fetchJoin() is used
• Applied in both AbstractJPAQuery (JPA) and AbstractHibernateQuery (Hibernate)
• Works across all JPA providers — no Hibernate-specific API dependency

Closes #1596

Problem

Since Hibernate 6, automatic result deduplication for fetch joins was removed, and Hibernate 7.3 fully removed the feature. This causes:

• fetch() returning duplicate parent entities from JOIN results
• fetchOne() throwing NonUniqueResultException when only one entity was expected

Solution

QueryDSL already knows when fetchJoin() is called (via JoinFlag.FETCH in query metadata). When detected, results are deduplicated using LinkedHashSet (preserves order) before returning.

This is applied at the QueryDSL level rather than using Hibernate's ResultListTransformer because:

• ResultListTransformer.uniqueResultTransformer() is Hibernate 7.3+ only
• A QueryDSL-level fix works across all JPA providers
• Normal queries without fetchJoin are unaffected

Test plan

• Existing tests pass (./mvnw -Pno-databases verify)
• No behavioral change for queries without fetchJoin
• Code formatted (pre-commit hook)