Fix official storage baseline reconciliation

[Rabithua]

What changed

• make official storage reconciliation skip stale object references that return 404 while still failing closed for unknown sizes and storage errors
• reconcile users in bounded batches with bounded object-store concurrency and prevent overlapping maintenance runs
• hide placeholder 0 B values until a user's baseline is authoritative
• clarify the role-exempt administrator storage label and omit zero-byte upload reservations

Root cause

The original baseline stopped on the first missing legacy object. Because the same earliest user was retried every minute, no user could finish reconciliation and the resource UI displayed a misleading zero value.

Impact

Existing physical objects are counted from authoritative HEAD responses. Missing legacy references consume no physical storage and no longer block the global baseline. Self-hosted unmanaged instances remain unchanged.

Validation

• Server resource tests: 9 passed
• Server TypeScript build
• Web tests: 161 passed
• Web production build
• ESLint for changed Server and Web files
• git diff --check