emrg: GUI i18n static-leak fix — wire 12 remaining data-i18n strings + regression tests

[argszero]

Summary

Completes rant 21:19 acceptance item "no hardcoded Chinese outside the i18n dict" — an audit found 12 static strings in index.html still hardcoded in Chinese with no data-i18n wiring, so English-system users see Chinese UI fragments. The i18n Stage 1/2 (#503/#504) migration missed them because they are not touched by JS at runtime (only updateGrowthCard/showEvolutionToast overwrite a subset).

Fixed leaks (index.html → data-i18n / data-i18n-title / data-i18n-placeholder)

1. growth-card title → copy.growthCardTitle
2. growth-note text → copy.growthNote (key existed, element unwired)
3. status-dot title → sidebar.statusTitle
4. back-to-bottom title → sidebar.backToBottom
5. model-form-name placeholder → settings.modelNamePlaceholder
6. Settings "关于" group title → settings.aboutTitle
7. Settings about-box hint → settings.aboutDesc
8. Settings "最近改进" group title → settings.recentTitle
9. evolution-toast title → copy.evolutionToastTitle (key existed, element unwired)
10. evolution-toast msg static → copy.evolutionToastMsgStatic (new non-parameterized key)
11. evolution-toast see button → copy.evolutionToastSee
12. evolution-toast dismiss button → copy.evolutionToastDismiss

New dict keys (zh+en, parity enforced by existing test)

copy.growthCardTitle, sidebar.statusTitle, sidebar.backToBottom, settings.aboutTitle, settings.aboutDesc, settings.recentTitle, settings.modelNamePlaceholder, copy.evolutionToastMsgStatic.

Regression tests (+2)

• index.html CJK scan: every line containing CJK must carry data-i18n/-title/-placeholder (or be an HTML comment) — catches exactly this leak class forever.
• New static key parity: spot-checks the new keys in en + zh.

Verification

• GUI suite: 84/84 pass (82 + 2 new). pytest 480 pass; import + emrg --help OK.
• Post-fix scan: 0 remaining CJK lines without data-i18n in index.html.

[argszero]

✅ LGTM — cycle 20260806-225341 (emrg-00c41753, author self-check, 1/3)

Verified the fix directly: (1) the leak scan script reports 0 remaining CJK lines without data-i18n in index.html (was 12); (2) each wired element is not overwritten by JS at runtime (checked app.js/dialogs.js — status-dot/back-to-bottom/model-form-name get className/value only), so the data-i18n title/placeholder actually renders localized; (3) toast msg static uses a non-parameterized key (copy.evolutionToastMsgStatic) so no literal {n} appears before showEvolutionToast overwrites it; (4) new keys are zh+en and the existing parity test enforces them; (5) GUI 84/84 + pytest 480 pass. This closes the last unmet acceptance item of rant 21:19.

[argszero]

✅ LGTM — cycle 20260806-230036 (emrg-00c41753, 2/3)

Verified the fix against master: exactly 12 Chinese static strings in index.html lack data-i18n (growth-card title/note, status-dot title, back-to-bottom, model-form-name placeholder, About title/desc, Recent-improvements title, evolution-toast title/msg/buttons) — all added by #501/#502 after i18n Stage 1 branched, so the migration missed them. The PR wires all 12 to data-i18n/title/placeholder with zh+en dict keys, and adds a leak-scan regression test (index.html Chinese must carry data-i18n) that would have caught this. This closes rant 21:19's "no hardcoded Chinese" acceptance item. CI Test green (run 31113853226).

[argszero]

✅ LGTM — cycle 20260806-231245 (emrg-00c41753, 3/3)

3 consecutive LGTMs from cycles 225341 / 230036 / 231245 — merging. Fix verified earlier: 12 index.html static Chinese strings wired to data-i18n with zh/en keys + index.html leak-scan regression test; CI green.