fix: reduce private checkpoint compression cost

[vincentkoc]

What Problem This Solves

Private checkpoint saves can spend several minutes on gzip compression, leaving too little time for authenticated database verification and publication.

User Impact

Use gzip level 1 to reduce packing CPU at the cost of larger transfers. Existing checkpoints remain readable. Full database validation, capacity limits and publication deadlines are unchanged; production timing remains unqualified.

Why This Change Was Made

The existing packing owner can select faster compression without adding workers, changing the checkpoint format or skipping validation. The production change is one parameter, with no net runtime LOC growth.

Evidence

• All 96 focused checkpoint Python tests pass, including authenticated restoration of both old level-9 and new level-1 assets with identical native database bytes, schema and counts. Existing corruption, partial-write, capacity and deadline tests pass.
• Independent source review and fresh P0–P3 autoreview are clean at 5d1684560ca3d3df719200e2bcf118db689b7982; syntax and diff checks pass.
• A bounded synthetic SQLite/FTS sample took 3.15s at level 9 and 0.08s at level 1, with a 16.7% larger compressed payload. This is a local tradeoff measurement, not a production speedup claim.
• All 20 hosted checks passed, including CI 39060079675 on the exact reviewed head and matching integration tree. A separately authorized manual end-to-end run must establish publication headroom before cadence activation. This PR does not activate a workflow or change acquisition scope.

[clawsweeper]

🦞👀
ClawSweeper picked this up.

Pull request received. I will update this pull request when review starts.