Issue transforming from PascalCase to kebab-case

Hi
Thanks for the library. I am trying to transform strings between PascalCase and kebab-case but there is a mistmatch when two consuctive charchters occure.

This is my code
```js
import * as changeCase from 'change-case';
const { pascalCase, kebabCase, splitSeparateNumbers } = changeCase;

const data = ['an-outlined-open-a-i', 'an-outlined-harmony-o-s', 'bx-sort-a-z'];

const pascalCases = data.map((icon) => {
return pascalCase(icon, {
mergeAmbiguousCharacters: true,
split: splitSeparateNumbers,
});
});

const kebabCasesFromPascal = pascalCases.map((icon) => {
return kebabCase(icon, {
split: splitSeparateNumbers,
});
});

console.log('Original Data:', data);
console.log('Pascal Cases:', pascalCases);
console.log('Kebab Cases from Pascal:', kebabCasesFromPascal);
```

which outputs
```
Original Data: [ 'an-outlined-open-a-i', 'an-outlined-harmony-o-s', 'bx-sort-a-z' ]
Pascal Cases: [ 'AnOutlinedOpenAI', 'AnOutlinedHarmonyOS', 'BxSortAZ' ]
Kebab Cases from Pascal: [ 'an-outlined-open-ai', 'an-outlined-harmony-os', 'bx-sort-az' ]
```

as you can see in Kebab Case from Pascal that `a-i` became `ai` and so on for the other strings. I disabled the options `mergeAmbiguousCharacters` and `splitSeparateNumbers` with the same results.

I am using version `5.4.4`

Reproduce the supplied example using the pascalCase and kebabCase entry points and compare the intermediate PascalCase values with the expected round trip. The payload names no source files or tests; first determine the intended behavior for consecutive characters such as “A-I”, then add regression coverage demonstrating the agreed result.

Written by the indexing model from the issue text.