[8.x] fix #37483 (aggregates with having)

[Tofandel]

The problem

Executing the following statement results in an sql error because the subquery is dismissed but the havings are kept:

User::query()->withCount('roles')->having('roles_count', '>', '2')->count()

Query:

select count(*) as aggregate from `users` having `roles_count` > 2

Error:

Illuminate\Database\QueryException with message 'SQLSTATE[42S22]: Column not found: 1054 Unknown column 'roles_count' in 'having clause' (SQL: select count(*) as aggregate from `users` having `roles_count` > 2)'

But executing the following runs without error

User::query()->withCount('roles')->having('roles_count', '>', '2')->get()

Query:

select `users`.*, (select count(*) from `roles` where `users`.`id` = `roles`.`user_id`) as `roles_count` from `users` having `roles_count` > 2

When comparing the queries of the previous statements, it is clear the original query should instead be wrapped within a subquery to be valid when dealing with havings to get a correct count (because having only applies to grouped data), exactly how it's currently done with unions, eg:

select count(*) as aggregate from (select `users`.*, (select count(*) from `roles` where `users`.`id` = `roles`.`user_id`) as `roles_count` from `users` having `roles_count` > 2) as `temp_table`

This PR

It applies the methodology of aggregates with union to aggregates with havings and does indeed fix the aforementioned issue #37483

It is not a breaking change because the previous behavior of this specific case wasn't previously working

TLDR

Queries with havings should behave exactly the same way as unions when executing aggregates functions

[driesvints]

@Tofandel thanks for the PR. It's best to add a much more thorough description to the PR so Taylor knows exactly why this is needed (see the PR description when submitting PR's).