Do not store GROUP BY clauses for Subqueries

[PaulWay]

A Subquery in an annotation on a queryset can contains its own aggregate functions (e.g. COUNT) on the subquery's fields. However, these subqueries do not need to be included as GROUP BY clauses in the original query.

[charettes]

Hello @PaulWay, I'm not sure if this is meant to address #30158 but as discovered in #11030 (comment) subquery annotations cannot always be removed from the group by clause.

For example, if a subquery annotation is explicitly included in a values() before an aggregate annotation the query must be grouped by it.

e.g.

def test_group_by_subquery_annotation(self):
    long_books_count_qs = Book.objects.filter(
            publisher=OuterRef('pk'),
            pages__gt=400,
        ).values(
            'publisher'
        ).annotate(count=Count('pk')).values('count')
    long_books_count_breakdown = Publisher.objects.annotate(
        long_books_count=Subquery(long_books_count_qs, IntegerField())
    ).values_list('long_books_count').annotate(
        total=Count('*'),
    )
    self.assertEqual(dict(long_books_count_breakdown), {
        None: 1,
        1: 4,
    })

[charettes]

Hey @PaulWay, after a bit more investigation I figured out that is a duplicate of an issue that is fixed in the upcoming Django 2.2 release (rc1 should be released this week) so I'm going to close this PR. Thanks for giving it a try!