refactor(web): reconcile sales:status* duplicate onto shared enums:status (#182)

[mforce]

Closes the "Reconcile the temporary sales:status* duplicate into a then-translated shared enums:status" follow-up tracked on #182.

What

The SalesPage status-filter dropdown (4 options) and the voided-payment badge rendered their labels from sales-local statusDraft/statusConfirmed/statusCancelled/statusVoided keys — a translated duplicate of enums:status.{Draft,Confirmed,Cancelled,Voided}. That duplicate existed only to avoid English-only labels back when enums wasn't translated; since enums joined TRANSLATED_NAMESPACES it's dead weight.

• SalesPage.tsx — 5 sites now render via statusLabel(...) (already imported and used elsewhere on the screen for the order-row badges). Option values (the server filter param) are untouched.
• en/es/tl — delete the 4 sales keys. Keep sales:allOption (filter-only "no status filter" chrome — no enums equivalent).

Behavior-preserving

enums:status values are byte-identical to the removed sales values in all three packs:

	Draft	Confirmed	Cancelled	Voided
en	Draft	Confirmed	Cancelled	Voided
es	Borrador	Confirmado	Cancelado	Anulado
tl	Draft	Kumpirmado	Kinansela	Na-void

Tests

• New wiring test: overriding enums:status.Confirmed flows through to the filter option text while its value stays the raw "Confirmed" code (mirrors the AuditPage precedent). Mutation-probed — hardcoding the label fails it.
• npm run test 1141 green · typecheck clean · build succeeds · i18n:scan still 3 (false positives) · catalogParity green (packs stay symmetric).

Scope

HistoryPage's separate history:status* set (its own vocabulary incl. statusSubmitted) is out of scope — the tracker item names sales specifically.

Not included

Unrelated local working-tree edits (deploy/docker-compose.yml, web/vite.config.ts) are deliberately not in this branch.

[mforce]

Review — unanimous approve (codex + 2 Claude agents + pi)

Reviewer	Verdict
codex	✅ approve — no correctness findings
pi	✅ approve — caveats were all verify-outside-diff, pre-satisfied
Claude (spec/correctness)	✅ approve — 8/8 checks green
Claude (adversarial)	✅ approve — no issues at confidence ≥ 80

Independently confirmed by reviewers:

• Byte-identity of enums:status.{Draft,Confirmed,Cancelled,Voided} vs the removed sales:status* in all three packs (en/es/tl) — no silent translation drift.
• Option value= attributes stay the raw status code (server filter param); only the child label changed.
• statusLabel imported and already used elsewhere on the screen; all four values are valid StatusValues (fallback path never hit).
• Wiring test is a real mutation-catcher, not a tautology: keySeparator:false makes "status.Confirmed" a literal key, so the override lands exactly where statusLabel reads; a hardcoded label or re-added local key would make the marker absent and the assertion throw.
• No other consumers of the deleted keys — HistoryPage's history:status* is a separate, out-of-scope namespace; Dashboard already uses statusLabel.
• Voided-payment badge: not a new semantic change — badge and dropdown already shared the same sales:statusVoided string; both now resolve to the byte-identical enums value.

Non-blocking suggestion (pi): a future multi-site/multi-locale test (voided badge + es/tl) would add breadth. The single wiring test + the removal's symmetry (caught by catalogParity) cover the regression surface for now.

Verified: npm run test 1141 green · typecheck clean · build ok · i18n:scan still 3 (false positives).