The doc layout puts a reading-time chip under the title on every doc page that has an entry in
`site.data["reading-time"]`, the fact the engine also writes to `reading-time.json`. The chip says
how long the page takes to read and, when enabled, how many words it has. A page with no entry, or
with a reading time of 0, gets no chip.

## Syntax

The layout adds it automatically. For a manual chip inside a page:

```markdown
:::reading-time{minutes=3 words=600 showWords}
:::
```

The numbers can also come from the page's data instead: `::reading-time{showWords}` shows the
page's own reading time and word count.

## Options

| Name | Type | Default | What it does |
| --- | --- | --- | --- |
| `minutes` | number | the page's reading time | The reading time in whole minutes. Overrides the stored value. |
| `words` | number | the page's word count | The word count. Overrides the stored value. |
| `showWords` | word | off | Show the word count next to the reading time. |

## Example

:::reading-time{minutes=3 words=600 showWords}
:::

:::reading-time{minutes=1}
:::

## Notes

- The chip reads `site.data["reading-time"]` by the page's exact URL, including `base`. Three
  shapes work: the engine's array of `{ url, words, minutes }`, the same list under `pages`, and a
  `pages` map keyed by the page URL. This lets a site override the numbers in
  `_data/reading-time.yml`.
- The text comes from `i18n/*.json` in the page's language. English distinguishes one minute from
  several; Persian uses one form for 0 and 1 and the same noun above that, since a Persian noun
  stays singular after a numeral; Arabic follows the CLDR zero, one, two, few (3 to 10), many
  (11 to 99), and other forms. The plural is chosen from the numeric value, never from its text.
- The visible minutes and words are written in the page language's digits — Persian `۵` and
  Arabic `٥`, English `5` — with the same numbering systems the calendar dates use, and regional
  tags like `fa-IR` or `ar-EG` inherit them. A supported `-u-nu-` preference on the language tag
  overrides the digits, for example `fa-IR-u-nu-latn` shows `5` where `fa-IR` shows `۵`, and
  `en-u-nu-arab` shows `٥` where `en` shows `5`; the wording and the plural category do not change.
  Large counts group by the page's own locale (`fa-IR` writes `۱٬۲۳۴٬۵۶۷`, and `en-IN` writes
  `12,34,567`), not the build machine's. A malformed tag or an unsupported preference falls back to
  Latin digits; nothing ever throws.
- `data-minutes` and `data-words` always keep the exact ASCII integers, whatever the visible
  digits look like, so other tooling can read the chip.
- A negative or unreadable `minutes` or `words` value falls back to the page's data, and a
  non-positive reading time renders nothing.