[9.x] Fix duplicated columns on select

[vdlp-mw]

This PR is based on a previous PR (#37936) that was not merged because there were objections to the proposed fix. This PR resolves those objections.

---

Using $query->addSelect('something') multiple times introduces duplicated columns, causing SQLSTATE[42S21]: Column already exists: 1060 Duplicate column name 'something'.

Steps to reproduce

Execute the following query:

// accounts() is a many-to-many relation, with an intermediate Eloquent Model
App\User::find(1)->accounts()->groupBy('accounts.id')->select('accounts.*')->paginate();

The resulting problematic SQL query will be:

select count(*) as aggregate
from (select `accounts`.*,
             `accounts`.*,
             `membership`.`user_id`    as `pivot_user_id`,
             `membership`.`account_id` as `pivot_account_id`,
             `membership`.`id`         as `pivot_id`,
             `membership`.`created_at` as `pivot_created_at`,
             `membership`.`updated_at` as `pivot_updated_at`
      from `accounts`
               inner join `membership` on `accounts`.`id` = `membership`.`account_id`
      where `membership`.`user_id` = 1
      group by `accounts`.`id`) as `aggregate_table`

Proposed fix

The previous PR was denied because it used array_unique on the columns array and may result in a different column order than expected. To prevent this, a strict in_array check is used to ensure column orders are not changed. The is_array check is there for sanity purposes as columns may not always be an array. This check can be dropped once strict types are introduced.

PR includes updated test to ensure the fix works. Test failed before fix and now passes so all should be fine.

Related issues

• PR: [8.x] Fix duplicated columns on select #37936
• Issue: Duplicate column name while aggregating #34096
• Discussion: Support for loose coupled duplicated column names for aggregators #34097