Aggregate FILTER may include rows with NULL predicates

**Describe the bug**

Native aggregation may include rows whose FILTER predicate evaluates to NULL, producing incorrect SUM and COUNT results.

**To Reproduce**

```sql
CREATE TABLE t_filter_null(amount INT, flag BOOLEAN) USING parquet;

INSERT INTO t_filter_null VALUES
(10, true), (100, NULL), (1, false), (200, NULL);

SELECT
SUM(amount) FILTER (WHERE NOT flag),
COUNT(*) FILTER (WHERE NOT flag)
FROM t_filter_null;
```

Auron returns `(301, 3)`, while Spark returns `(1, 1)`.

**Expected behavior**

Only rows whose FILTER predicate is non-null and true should contribute to the aggregate.

**Screenshots**

**Additional context**

Start by running the SQL reproduction in the issue and compare Auron's results with Spark's. Trace native aggregate FILTER handling for NULL predicates; done when SUM and COUNT include only rows where the predicate is non-null and true, matching the expected (1, 1) result.

Written by the indexing model from the issue text.