Describe the bug

Extracted from a discussion on the equals kernels

It is currently very easy to get inconsistant state between the nullable flag on StructArray fields and the presence of their validity bitmap. This can lead to very difficult to debug issues, as for example the equality comparison of StructArray includes this flag, but it is not printed by the Debug impl, for example in assertions.

To Reproduce

Example of a confusing test:

    #[test]
    fn compare_nullable_struct_array_fields() {
        let array1 = StructArray::try_from(vec![(
            "a",
            Arc::new(Int32Array::from(vec![1, 2, 3])) as ArrayRef,
        )])
        .unwrap();
        let array2 = StructArray::try_from(vec![(
            "a",
            Arc::new(Int32Array::from(vec![Some(1), Some(2), Some(3)])) as ArrayRef,
        )])
        .unwrap();
        // assert_eq!(array1.data_type(), array2.data_type());
        assert_eq!(&array1, &array2);
    }

panicked at 'assertion failed: `(left == right)`
  left: `StructArray
[
-- child 0: "a" (Int32)
PrimitiveArray<Int32>
[
  1,
  2,
  3,
]
]`,
 right: `StructArray
[
-- child 0: "a" (Int32)
PrimitiveArray<Int32>
[
  1,
  2,
  3,
]
]`'

Expected behavior

• if a field is not nullable it should not have a validity bitmap
• a StructArray itself should only have a validity bitmap if all its fields are nullable
• maybe StructArray::from(Vec<(&str, ArrayRef>) should always set the nullable flag?

A clear and concise description of what you expected to happen.

Additional context

Alternatives could be to ignore the nullable flag when comparing arrays