[import/order] Regression in `type` behavior with `parent` and `sibling`

[erunion]

We have a complex import/order config and are seeing some new behavior with our type imports on v2.27.5 where previously sorted parent → sibling type imports want to now be sorted as sibling → parent.

'import/order': [
  'error',
  {
    'alphabetize': {
      order: 'asc',
      caseInsensitive: true,
    },
    'groups': ['type', 'builtin', 'external', 'internal', 'parent', 'sibling', 'index', 'object'],
    'newlines-between': 'always',
    'warnOnUnassignedImports': true,
  },
],

Before

import type { CommandOptions } from '../lib/baseCommand';
import type { Options } from './openapi/validate';

import Command from '../lib/baseCommand';
import isHidden from '../lib/decorators/isHidden';

import OpenAPIValidateCommand from './openapi/validate';

After

import type { Options } from './openapi/validate';
import type { CommandOptions } from '../lib/baseCommand'; <-- this line was moved 

import Command from '../lib/baseCommand';
import isHidden from '../lib/decorators/isHidden';

import OpenAPIValidateCommand from './openapi/validate';

[ljharb]

I think this is a duplicate of #2682?

[erunion]

@ljharb I'm not sure because the non-type imports are fine here, it's just import type that's affected.

[ljharb]

cc @Pearce-Ropion in case this is related, but if not, it'd still be great to tackle it in the same PR

[Pearce-Ropion]

Hmm, this is caused by the same alphabetical sorting issue but cannot be fixed in the same way unless we add groups for types.

[ljharb]

types is already its own group, and its contents should have the top-level groups sorting, minus types, applied.

[Pearce-Ropion]

Quick progress update, I have a fix for this working locally. I'm working on adding a few more unit test cases to maintain code coverage. I should have a PR up by the end of the week.

[Pearce-Ropion]

I've put up #2721. I think it needs more unit testing but figured I'd get the PR out there