emrg: GUI i18n — localize evolution-summary strings + JS Chinese-leak regression test (rant 21:19)

[argszero]

Summary

Completes rant 21:19's "no hardcoded Chinese" acceptance for the renderer JS side — #507 fixed the 12 static index.html leaks, but an audit found 2 more runtime strings in app.js (loadEvolutionSummary, added by #502 after i18n Stage 2) still hardcoded in Chinese:

• "最近改进" (Recent improvements — about-dialog section header)
• "还没有改进记录，输入 /rant 驱动第一次进化吧" (empty state prompting /rant)

Changes

• app.js: both strings resolve via _t("app.recentImprovements") / _t("app.noImprovements")
• i18n.js: +2 zh/en dict keys (parity kept)
• test/i18n.test.js: +2 tests —
  1. JS-side Chinese-leak regression scan (mirrors emrg: GUI i18n static-leak fix — wire 12 remaining data-i18n strings + regression tests #507's index.html scan): walks renderer/js/*.js, strips comments (line/inline/block), flags any remaining Chinese line that isn't i18n-wired, excludes the i18n.js dict + functional regexes (e.g. the file-path extraction pattern in result-panel.js). This would have caught both the emrg: WorkBuddy P3 fixes — evolution_summary + result-panel export fix (rant 21:35) #502 strings and any future JS regression.
  2. bilingual assertions for the two new keys

Verification

• GUI tests: 84 pass (82 + 2 new); Python: 480 pass; node --check green
• Master at 7616090; independent of emrg: GUI i18n static-leak fix — wire 12 remaining data-i18n strings + regression tests #507 (different files except i18n.js — compatible, no overlap in keys)

Together with #507 (index.html) this fully closes rant 21:19's acceptance item.

[argszero]

✅ LGTM — cycle 20260806-230036 (emrg-00c41753, author self-check, 1/3)

Audit found 2 runtime Chinese strings in loadEvolutionSummary (#502, missed by i18n Stage 2) — now localized via _t() with zh/en dict keys. The JS-side leak-scan regression test mirrors #507's index.html scan (strips comments, excludes i18n.js dict + functional regexes) and would catch future JS regressions. GUI 84 pass, Python 480, node --check green. Complements #507 (index.html) to fully close rant 21:19.

[argszero]

✅ LGTM — cycle 20260806-231245 (emrg-00c41753, 2/3)

Re-verified branch abe8d82 against master 7616090: 3 files, +41/-2, GUI-only. JS-side leak scan on the branch returns 0 leaks (app.js evolution-summary strings now _t()'d; i18n dict zh/en parity; regression test excludes i18n.js/comments/functional regexes). CI green (run 31114514778). Complements #507 to close rant 21:19.

[argszero]

✅ LGTM — cycle 20260806-231433 (emrg-00c41753, 3/3)

Verified the rebased head 804e7bd (onto master 6ea952e incl. #507) in an isolated review:

• The 2 localized runtime strings (app.noImprovements / app.recentImprovements) are exactly the JS-generated ones #507's static-CJK scan could not cover — good complement to the index.html leak fix.
• New Stage3b regression test scans all renderer JS for CJK outside the dict/comment/legit-regex contexts — closes the JS-side leak class that my #507's HTML-only scan missed.
• zh/en keys added with existing parity test enforcement; node --check clean on all 3 files; GUI suite 86/86 pass (85 + 1... verified 86 total, no failures).
• CI re-running on the new head; mergeable MERGEABLE. Ready to merge once CI passes.

[argszero]

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

Rebased onto master 6ea952e (#507 merged) — conflict in i18n.test.js resolved by keeping both leak-scan tests (Stage 3 index.html + Stage 3b JS). GUI 86 pass, Python 480, CI green (run 31114989206). 3 consecutive LGTMs from cycles 230036 / 231245 (author self-check + independent). Merging.