arrow-arith is_null is buggy with NullArray

[smiklos]

Describe the bug

When using the arith kernel's is_null or is_not_null on arrays of type NullArray (with datatype Null) the .nulls() buffer is always empty for these types. Therefor the kernel returns an array with all values false/true which is the exact opposite of what should be returned.

To Reproduce

Noticed it here apache/datafusion#7038

Run this unit test in datafusion/is_null.rs

   fn is_null_op_allnull() -> Result<()> {
        let schema = Schema::new(vec![Field::new("a", DataType::Null, true)]);
        let mut a = NullBuilder::with_capacity(1);

        // expression: "a is null"
        let expr = is_null(col("a", &schema)?).unwrap();
        let batch = RecordBatch::try_new(Arc::new(schema), vec![Arc::new(a.finish())])?;

        let result = expr.evaluate(&batch)?.into_array(batch.num_rows());
        let result =
            as_boolean_array(&result).expect("failed to downcast to BooleanArray");

        let expected = &BooleanArray::from(vec![true]);

        assert_eq!(expected, result);

        Ok(())
    } 

Expected behavior

for is_null I expect a BoolArray of true values for NullArray input (irrespective of length).
for is_not_null I expect a BoolArray of false values for NullArray input (irrespective of length).

Additional context

[smiklos]

I can create a pr for this with my solution but was curious if the NullArray::nulls should return a nullbuffer based on self length or if the kernel should be made Null aware

[tustvold]

I think the kernel should special case NullArray, as the arrow specification states that a NullArray should not have a null buffer

[tustvold]

label_issue.py automatically added labels {'arrow'} from #4566