[Issue #1415] Retire fully deleted GC groups with checkpoint protection

[AntiO2]

Fully deleted Storage GC groups currently delete their empty output and return without retiring the old files. Repeated updates therefore leave obsolete files permanently REGULAR.

Retire these groups atomically without a replacement file. Journal their safe visibility-fold timestamp, confirm retirement only after a covering recovery checkpoint, protect pending journal files from physical cleanup, and restore the old catalog if the selected checkpoint predates retirement. Equal-invalidity candidates now prefer older file IDs so a bounded group limit does not repeatedly select newly created groups before older ones.

Validation:

• With parent Surefire skipTests temporarily set to false: mvn -pl pixels-retina -am -Dtest=TestStorageGcWal,TestEmptyStorageGc -Dsurefire.failIfNoSpecifiedTests=false test: 20 tests passed, none skipped.
• JUnitCore selected atomicSwapFiles tests: 5/5 passed, including retirement without promotion and rejected empty requests. The daemon's default test provider discovers no JUnit 4 tests, so these were invoked explicitly.
• Full DAO JUnitCore suite: 19/21; two insertBatch fixture failures were reproduced against unmodified master and are unrelated to this change.
• Isolated 13-minute closed-loop replay with two-minute retention completed: 26,506,764 row events / 138 complete cycles, all eight tables recovered their initial active totals (870,600,000 overall), and 18 old files / 124,220,282 bytes were physically deleted. This validates retirement and cleanup only; it used a diagnostic native mutex and is not a final performance or steady-state measurement.
• Integrated 15-minute validation without the diagnostic native mutex completed 10 partial-live rewrites, restored all table populations, and physically deleted 28 files / 125,261,498 bytes. Two-minute retention and warm-up pauses make this a correctness run only.

Closes #1415

Additional combined live validation after the separate sink same-key coalescing fix (pixelsdb/pixels-sink#22): HyBench SF100 original Proto replay ingested 22,945,933 rows and fully committed all 452,616 admitted transactions. Two GC cycles and three partial-live group rewrites completed. After draining, savingaccount/checkingaccount each had 30,200,000 active versions, customer 30,000,000 and company 200,000, exactly matching their initial populations; insert-growing tables are reported separately. Sink shutdown had no error entries. This is functional integration evidence, not a steady-state or isolated performance comparison. The repeated formal matrix has resumed. Results: gengdy1545/retina-graph#34.