Extract shared SQL user access with existing backend adapters

[BenCodez]

Dependency and merge order

Depends on #320. Merge FLAT removal first. This branch contains #320's commit followed by one SQL-access commit. It targets master so the existing Maven PR workflow validates the combined candidate without adding or modifying a workflow. The incremental SQL-only diff is e0f56679b12b0f0d7e67c655383e0f33b628d938..b789d7b1b95252f05083f717a039fc9ff0841975 (10 files).

Summary

• Add platform-neutral core.user.storage.SqlUserStorage and SqlUserDataAccess for live SQL row interpretation, typed reads/writes, keys/values, existence, deletion, and bulk writes.
• Add BukkitSqlUserStorage adapting the existing SQLite/MySQL user-table providers through lazy plugin/UUID suppliers. No new connection pool, JDBC implementation, cache, identity resolver, executor, or write queue.
• Make the existing UserData SQL paths use the shared access layer immediately. Keep its public method signatures and private temporary-cache field/accessors.
• Preserve all six fetch modes, cache precedence, notifications, queue short-circuiting, execution-time UUID lookup, list encoding, and existing write scheduling.
• Preserve virtual getMySqlRow, getSQLiteRow, and convert hooks for existing subclasses. Readers do not capture a stale row/provider during construction.
• Add 14 JUnit tests plus a headless behavioral fixture: typed/default/case behavior, duplicate handling, mutable exports, lazy ownership, SQL delegation, bulk call ordering, facade overrides, all fetch modes, queued/async writes, failure handling, and isolated loading without Bukkit/plugin classes.

Compatibility and scope

This is the SQL access boundary, not a claim that the entire native storage/user runtime is finished. Bukkit-facing provider construction, SQLite bootstrap, schema registration, UserManager enumeration, user-cache lifecycle, and final native packaging remain follow-ups. Native code can supply a storage port without an AdvancedCorePlugin; the current Bukkit adapter retains the existing providers.

The provider's existing behavior is retained, including the MySQL unavailable-provider bulk no-op and SQLite cumulative per-entry bulk updates. No new durable-completion, atomic-batch, or retry guarantee is invented. #317's async reward checkpoints/completion semantics remain separate and must be preserved when reconciling overlapping work.

One Maven project. No POM, dependencies, workflows, schema, persistence encoding, migration, or reward-execution changes beyond the prerequisite FLAT removal. SimpleAPI stays 1.0.2-SNAPSHOT; #314 stays closed and the unwanted pinning test stays removed.

Verified GitHub Actions build

Java CI with Maven — run 603 passed. Inspected completed build job 103632505558 and its Maven log:

• mvn -B -f AdvancedCore/pom.xml package on Temurin Java 21.0.12.
• 351 tests run, 0 failures, 0 errors, 0 skipped.
• All 14 new tests passed: BukkitSqlUserStorageTest (3), SqlUserDataAccessTest (5), SqlUserDataFacadeTest (5), and SqlAccessHeadlessTest (1).
• The headless fixture executed shared SQL-access behavior with Bukkit/plugin classes denied. Its in-memory storage port is not a real database integration test.
• The prerequisite's SQL-only tests and existing user/cache/fetch-mode/reward/lifecycle/platform regressions passed.
• Maven refreshed snapshot metadata and downloaded SimpleAPI 1.0.2-20260912.212854-22. This records tested inputs; the POM remains 1.0.2-SNAPSHOT.
• AdvancedCore/target/AdvancedCore.jar generated; shading/minimization completed; BUILD SUCCESS.
• Tested integration merge 248544428ea8d3b9d4d120390cc3f8bdf635c155 for candidate b789d7b1b95252f05083f717a039fc9ff0841975 and base b03cb7f8a26c22e1e35b0c1a44ac1df7d5d8098d.

These are GitHub Actions results, not full local Maven validation. Shade emitted overlap/minimization warnings; successful packaging does not establish packaged runtime compatibility.

Local checks

• Syntax-parsed the 15 Java files present in the combined local partial worktree with JDK 21; nine are changed/added by this SQL commit. This is not dependency/type checking.
• Mechanically compared the existing UserData public method signatures: unchanged. Both scalar cache-selection prefixes match the prerequisite exactly; the original tempCache field is retained.
• git diff --cached --check passed for the complete 10-file incremental diff in an exact-baseline partial Git fixture. The uploaded UserData blob matches the locally checked bytes.
• GitHub comparison confirms one incremental commit, 10 files, 771 additions and 180 deletions beyond Remove deprecated FLAT user storage #320.

Draft validation limits

Maven is absent and external dependency/repository DNS is unavailable in the editing container. Full local/downstream builds, real database round trips, packaged linkage, live server testing, and a fresh independent reviewer have not run. Same-context source inspection is not an independent clean verdict.

Prerequisite: e0f56679b12b0f0d7e67c655383e0f33b628d938 (#320).
Candidate: b789d7b1b95252f05083f717a039fc9ff0841975.

No merge, release, deployment, or manual external review request was performed.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-12T22:58:47.678301Z	b789d7b	Draft marked ready
🔒 Security Review	✅ Completed 2026-09-12T23:01:21.013476Z	b789d7b	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.