`SELECT DISTINCT` returns duplicate rows for statistical aggregates grouped by an unprojected key

MariaDB returns duplicate rows from a `SELECT DISTINCT` query when a
statistical aggregate is projected and the `GROUP BY` contains an additional
unprojected key.

Each group in the reduced example has the same variance, date, and projected
modulo value. The groups differ only by `g`, which is intentionally retained as
a grouping key but is not part of the result. The two projected rows are
therefore identical and `DISTINCT` must collapse them into one row. MariaDB
12.3.2 instead returns both copies.

Moving the grouped relation behind a derived table and applying `DISTINCT` in
the outer query returns the correct single row. The issue also affects
`STDDEV_SAMP`, `VAR_POP`, and `STDDEV_POP` in the same query shape.

Two independently generated VECT queries exposed this same defect. One query
using `STD` (an alias of `STDDEV_POP`) returned 64 copies of one completely
identical row, although `DISTINCT` should have returned one row. A second query
using `VAR_SAMP` returned 100 rows containing only 89 distinct projected rows;
11 rows should have been eliminated. Both reduce to the query shape below, so
they are treated as two triggers for one database bug rather than as separate
bugs.

1. How to repeat

The following setup is self-contained:

```sql

DROP DATABASE IF EXISTS distinct_statistical_aggregate_test;

CREATE DATABASE distinct_statistical_aggregate_test;

USE distinct_statistical_aggregate_test;

CREATE TABLE samples (

    g INT,

    x INT,

    d DATE

);

INSERT INTO samples VALUES

    (1, 10, '2025-01-01'),

    (1, 10, '2025-01-01'),

    (2, 20, '2025-01-01'),

    (2, 20, '2025-01-01');

```

Run the direct `DISTINCT` query:

```sql

SELECT DISTINCT

       VAR_SAMP(x) AS variance_value,

       d,

       MOD(g, g) AS modulo_value

FROM samples

GROUP BY d, MOD(g, g), g;

```

MariaDB 12.3.2 returns two identical rows:

```text

+----------------+------------+--------------+

| variance_value | d          | modulo_value |

+----------------+------------+--------------+

|              0 | 2025-01-01 |            0 |

|              0 | 2025-01-01 |            0 |

+----------------+------------+--------------+

```

Apply `DISTINCT` outside the grouped relation as a control:

```sql

SELECT DISTINCT variance_value, d, modulo_value

FROM (

    SELECT VAR_SAMP(x) AS variance_value,

           d,

           MOD(g, g) AS modulo_value

    FROM samples

    GROUP BY d, MOD(g, g), g

) AS grouped_result;

```

The control query correctly returns one row:

```text

+----------------+------------+--------------+

| variance_value | d          | modulo_value |

+----------------+------------+--------------+

|              0 | 2025-01-01 |            0 |

+----------------+------------+--------------+

```

The direct query also returns duplicate rows when `VAR_SAMP` is replaced
with any of the following expressions:

```text

STDDEV_SAMP(x)

VAR_POP(x)

STDDEV_POP(x)

```

`AVG(x - x)` returns the correct single row in the same query shape.

1. Expected result

The direct and derived-table forms should both return one row. Although the
grouped relation has two groups, their projected values are identical, so
`SELECT DISTINCT` must eliminate the duplicate:

```text

0, 2025-01-01, 0

```

1. Actual result

MariaDB 12.3.2 returns two identical rows from the direct query. Applying the
same `DISTINCT` operation outside a derived table returns the correct single
row.