Port upstream CodexBar 0.52.0

[Finesssee]

Port of upstream steipete/CodexBar v0.52.0 (stacked on #344; read-only consumption of tagged sources, zero upstream contact per docs/PORTING.md).

Ported

Item	Notes
Usage & Spend per-project aggregation (steipete#2984)	usage_spend.rs: UsageSpendModelRow struct + extend_model_rows() joins 7d/30d model-level cost breakdowns into the spend summary. UsageSpendTab.tsx: ModelsPanel renders per-model rows with show-all/show-less toggle. ProjectsPanel renders ranked Codex projects with expandable session lists.
Usage & Spend Projects panel + expand (steipete#2985)	UsageSpendTab.tsx: ProjectsPanel with expandedProjects Set state, per-project expand/collapse showing sessions filtered by projectId. Show-all/show-less for >8 projects.
Cost history chart localization (steipete#2983)	Locale keys UsageSpendModels, UsageSpendProjects, UsageSpendShowAll, UsageSpendShowLess, UsageSpendConversations, UsageSpendPartialHistory, UsageSpendNoModels, UsageSpendNoProjects added to keys.ts, locale.rs, and en-US.ftl.
Grok SuperGrok Heavy from CLI settings (steipete#2991)	grok/mod.rs: fetch_cli_subscription_tier() calls cli-chat-proxy.grok.com/v1/settings with 2s timeout, grok_plan_display_name() normalizes subscription_tier_display to "SuperGrok Heavy"/"SuperGrok". Browser path falls back to local credentials for identity enrichment only.

Skipped (no Windows counterpart — macOS-only)

Item	Reason
Mission Control keepalive window removal (steipete#2955, steipete#2997, steipete#3003)	No Mission Control / Spaces on Windows. Tauri manages window lifecycle directly.
Retained NSWindow controller fix (steipete#3029)	macOS AppKit window controller pattern. Tauri webview windows have no equivalent.

Deferred (evidence)

Item	Reason
Full upstream 0.52 locale parity for all 23 languages	Only en-US ported. Other 6 locale catalogs (es-MX, ja-JP, ko-KR, ru-RU, zh-CN, zh-TW) need translation pass.

Verification

• cargo fmt --all — clean
• cargo clippy --manifest-path rust/Cargo.toml --all-targets -- -D warnings — clean
• cargo clippy --manifest-path apps/desktop-tauri/src-tauri/Cargo.toml --all-targets -- -D warnings — clean
• cargo test --manifest-path rust/Cargo.toml --lib — passed
• cargo test --manifest-path apps/desktop-tauri/src-tauri/Cargo.toml — passed
• pnpm --dir apps/desktop-tauri test — passed
• pnpm --dir apps/desktop-tauri run build — passed
• scripts/local-check.ps1 — not run: known StrictMode bug in the script causes terminating errors on undefined variables. Documented as gate limitation.
• CUA Windows desktop proof — pending: see thermo-nuclear review comments. UI surfaces changed (UsageSpendTab, TrayPanel).

Thermo-nuclear review

A thermo-nuclear code-quality review was performed. Blocking: this PR introduces UsageSpendModelRow/extend_model_rows/ModelsPanel which PR #346 immediately deletes and replaces with SpendContract. These PRs should be collapsed. See review comment: #345 (comment)

[Finesssee]

Thermo-nuclear review: BLOCKING.

PR #345 introduces UsageSpendModelRow, extend_model_rows(), model sorting logic, ModelsPanel, and ProjectsPanel (~200 lines of backend + frontend model aggregation). PR #346 then deletes every line of that and replaces it with SpendContract + ContractModelsPanel + SpendContractOverview + ActivityHeatmap + MetricCard.

This is stacked churn. If you merge #345 first, you ship a structure you immediately throw away in #346. The two PRs should be collapsed: either squash #345 into #346 (the SpendContract is the real design), or #345 should introduce SpendContract from the start instead of the throwaway UsageSpendModelRow intermediate.

Action: Collapse #345 and #346 into a single PR that introduces SpendContract directly.

[Finesssee]

Closing per thermo-nuclear review: the 0.52 intermediate UsageSpendModelRow/ModelsPanel structure is intentionally removed by the canonical SpendContract design in the next stack layer. PR #346 will be retargeted to port/upstream-0.51-review so 0.52+0.53 land as one coherent spend architecture without stacked throwaway churn.