i18n: generate zh-Hant from zh-CN

zh-Hant is the zh-CN text in Traditional characters, but contributors
still translated it by hand from `--missing zh-Hant`, and a model asked
for Traditional Chinese writes Taiwan vocabulary (影片, 伺服器, 載入).

`npm run i18n -- --zh-hant` now rewrites zh-Hant.json from zh-CN.json
with OpenCC s2tw (opencc-js 1.4.2, cn -> tw, characters only), then the
fixed list of characters it picks wrongly. The check that `npm test`
runs fails when zh-Hant differs from that conversion, naming the key,
and `--missing zh-Hant` points to the command instead of printing a
template.

The conversion reproduces every one of the 3346 entries on main. 25
keys were out of zh-CN order and move; no value changes.

The converter loads on first use, so a copy of the script without
node_modules (stale-index.test.mjs) still extracts messages.

docs/TRANSLATORS.md

@@ -36,6 +36,8 @@ npm run i18n -- --prune           # drop entries the code no longer uses
 
 Changing English text changes its key: the old translation becomes unused and the new text shows as missing. Translate it and prune.
 
+Do not translate `zh-Hant`: it is generated. Translate `zh-CN`, then run `npm run i18n -- --zh-hant` (see [zh-Hant.md](./translations/zh-Hant.md)).
+
 ## Fixing a translation
 
 Edit `src/i18n/locales/<code>.json` and run `npm run i18n`. To see your change, open the app with `?lang=<code>`.