Add sync-ready quota history for Codex

[starkdmi]

Summary

• Collect Codex quota snapshots from token_count traces during normal and archive scans.
• Store content-addressed raw payloads, normalized quota observations, account attribution, and optional usage-event links.
• Reconstruct quota reset cycles with percentage change points, reset drift tolerance, overlap warnings, and deduplicated usage totals.
• Add quota status, current, windows, history, and export commands.
• Publish a standalone multi-device sync projection for attributed windows of at least seven days without changing or transmitting through SyncBatch.v3.
• Add regression coverage for rescans, attribution, copied traces, usage linking, reconstruction, projections, CLI output, and forward-schema compatibility.

Notes

• Historical observations remain unassigned unless covered by a timestamped source-account assignment.
• Unassigned observations are partitioned by local source to prevent different accounts from being merged.
• Raw provider payloads, local source identities, tokens, and costs are excluded from sync projections.
• Server-side token and cost totals are intended to come from existing deduplicated usage rollups.

Validation

• cargo build --workspace
• cargo test --workspace
• cargo fmt --check
• Clippy with warnings denied
• git diff --check
• Targeted quota CLI and forward-schema tests