interval analysis error when have two filterexec that inner filter proves zero selectirity

[haohuaijin]

Describe the bug

When two FilterExec nodes are stacked and the inner filter proves zero selectivity (no rows can match), the outer filter panics during interval analysis.

Root cause: When a FilterExec determines that no rows can pass its predicate (e.g., a > 200 when a's max is 100), collect_new_statistics produced column statistics with untyped ScalarValue::Null for min/max/sum values. The Null variant has data type Null.

If an outer FilterExec sits on top and tries to analyze its own predicate (e.g., a = 50), it attempts to intersect intervals from the inner filter's statistics (Null type) with the literal in its predicate (Int32 type). Interval::intersect requires both sides to have the same data type, so it panics with:

"Only intervals with the same data type are intersectable, lhs:Null, rhs:Int32"

To Reproduce

    #[tokio::test]
    async fn test_nested_filter_with_zero_selectivity_inner() -> Result<()> {
        // Inner table: a: [1, 100], b: [1, 3]
        let schema = Schema::new(vec![
            Field::new("a", DataType::Int32, false),
            Field::new("b", DataType::Int32, false),
        ]);
        let input = Arc::new(StatisticsExec::new(
            Statistics {
                num_rows: Precision::Inexact(1000),
                total_byte_size: Precision::Inexact(4000),
                column_statistics: vec![
                    ColumnStatistics {
                        min_value: Precision::Inexact(ScalarValue::Int32(Some(1))),
                        max_value: Precision::Inexact(ScalarValue::Int32(Some(100))),
                        ..Default::default()
                    },
                    ColumnStatistics {
                        min_value: Precision::Inexact(ScalarValue::Int32(Some(1))),
                        max_value: Precision::Inexact(ScalarValue::Int32(Some(3))),
                        ..Default::default()
                    },
                ],
            },
            schema,
        ));

        // Inner filter: a > 200 (impossible given a max=100 → zero selectivity)
        let inner_predicate: Arc<dyn PhysicalExpr> = Arc::new(BinaryExpr::new(
            Arc::new(Column::new("a", 0)),
            Operator::Gt,
            Arc::new(Literal::new(ScalarValue::Int32(Some(200)))),
        ));
        let inner_filter: Arc<dyn ExecutionPlan> =
            Arc::new(FilterExec::try_new(inner_predicate, input)?);

        let outer_predicate: Arc<dyn PhysicalExpr> = Arc::new(BinaryExpr::new(
            Arc::new(Column::new("a", 0)),
            Operator::Eq,
            Arc::new(Literal::new(ScalarValue::Int32(Some(50)))),
        ));
        let outer_filter: Arc<dyn ExecutionPlan> =
            Arc::new(FilterExec::try_new(outer_predicate, inner_filter)?);

        let statistics = outer_filter.partition_statistics(None)?;
        assert_eq!(statistics.num_rows, Precision::Inexact(0));

        Ok(())
    }

Expected behavior

work without panic

Additional context

No response

[haohuaijin]

take