sliceAnsi returns wider result than specified endColumn for wide characters

sliceAnsi returns wider result than specified endColumn for wide characters

## Description

`sliceAnsi` returns a result wider than the specified `endColumn` when the boundary falls inside an East Asian wide (CJK) character. The character is included in full ("rounded up"), so the result exceeds the requested column range.

## Reproduction

```js
import sliceAnsi from 'slice-ansi';
import stringWidth from 'string-width';

const text = 'あいう'; // each char is 2 columns wide

for (let end = 0; end <= 6; end++) {
  const s = sliceAnsi(text, 0, end);
  console.log(`end=${end} → ${JSON.stringify(s)} (width=${stringWidth(s)})`);
}
```

### Output

```
end=0 → "" (width=0)
end=1 → "あ" (width=2)  ← expected width ≤ 1
end=2 → "あ" (width=2)
end=3 → "あい" (width=4) ← expected width ≤ 3
end=4 → "あい" (width=4)
end=5 → "あいう" (width=6) ← expected width ≤ 5
end=6 → "あいう" (width=6)
```

### Expected

When `endColumn` falls in the middle of a wide character, the character should be excluded so that `stringWidth(result) <= endColumn - startColumn`. This is consistent with the API doc: "Zero-based visible-column index at which to end the slice."

### Actual

The wide character is included in full, causing the result to be 1 column wider than the requested range.

## Impact

This causes downstream bugs in packages like `cli-truncate` (sindresorhus/cli-truncate#28), where truncated output exceeds the specified `maxWidth` for CJK text.

## Environment

- slice-ansi 7.1.0
- string-width 7.2.0
- Node.js v25.8.2