Feat: Add Codex tiered and priority pricing

[iam-brain]

Summary

• add Codex priority trace enrichment from local appserver logs_2.sqlite without persisting websocket request bodies
• apply Codex base pricing per rollout token row, then add priority/Fast surcharge only for matched priority rows at <=272,000 input tokens
• long-context Codex pricing for any priority-marked row above the 272k input boundary
• gate SQLite-backed trace scanning/tests on canImport(SQLite3), falling back to base pricing on platforms without SQLite3
• bump the Codex cost cache artifact and cover priority parsing, privacy, missing-DB fallback, per-row pricing, cumulative-total, cached-input threshold, and long-context regressions

Validation

• swift test --filter CostUsageScannerCodexPriorityTests passed: 5 tests
• swift test --filter CostUsageScannerPriorityTests passed: 5 tests
• swift test --filter CostUsagePricingTests passed: 23 tests
• make check passed: 0 SwiftFormat/SwiftLint violations
• earlier full focused suite passed: swift test --filter "CostUsagePricingTests|CostUsageScannerPriorityTests|CodexPriorityTraceScannerTests|CostUsageCacheTests|ModelsDevPricingTests" passed: 51 tests
• swift build -c release --product CodexBarCLI passed
• ./Scripts/compile_and_run.sh built the app target
• validated against local ~/.codex/logs_2.sqlite and rollout JSONL with swift run CodexBarCLI cost --provider codex --format json --pretty --refresh

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 6394cc9012

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Persist the Codex turn across incremental scans

When scanCodexFile resumes from a cached parsedBytes offset, parseCodexFile restores initialModel and initialTotals but this new currentTurnID always starts as nil. If CodexBar scans while a priority turn is still being appended, the cached prefix may already contain the task_started event and later token_count rows commonly rely on that state instead of repeating turn_id; those newly appended rows then get cached with turnID == nil, so codexPrioritySurchargeUSD never applies priority pricing for the rest of that turn until a full rescan.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 70c6487f85

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Include adjacent days when loading priority turn metadata

loadCodexDaily fetches priority turn metadata only for range.sinceKey...range.untilKey, but Codex token rows can land on a new day while still belonging to a turn that started just before midnight. In that cross-day case, rows for the report day keep the prior turn ID, but priorityTurns omits that turn, so Fast/Priority surcharge is silently skipped for those rows. This under-reports cost for overnight sessions and can be fixed by loading priority metadata over the scanner window (scanSinceKey/scanUntilKey) or another boundary-safe span.

Useful? React with 👍 / 👎.