DecimalArray::from_fixed_size_list_array fails when offset > 0

[HaoYang670]

Describe the bug
from_fixed_size_list_array cannot return a correct result when array.offset() > 0

To Reproduce

    #[test]
    fn test_decimal_array_from_fixed_size_list() {
        let value_data = ArrayData::builder(DataType::UInt8)
            .len(48)
            .add_buffer(Buffer::from_slice_ref(&[12_i128, 34, 56]))
            .build()
            .unwrap();
        
        let null_buffer = Buffer::from_slice_ref(&[0b_101]);

        // Construct a list array from the above two
        let list_data_type = DataType::FixedSizeList(
            Box::new(Field::new("item", DataType::UInt8, false)),
            16,
        );
        let list_data = ArrayData::builder(list_data_type.clone())
            .len(2)
            .null_bit_buffer(Some(null_buffer))
            .offset(1)
            .add_child_data(value_data.clone())
            .build()
            .unwrap();
        let list_array = FixedSizeListArray::from(list_data);
        let decimal = DecimalArray::from_fixed_size_list_array(list_array, 38, 0);

        assert_eq!(decimal.len(), 2);
        assert!(decimal.is_null(0));
        assert_eq!(decimal.value_as_string(1), "56".to_string());
    }

Expected behavior
Pass the test.

Additional context
The reason of the failure is that we drop the offset when copying value_buffer and null_buffer:
https://github.com/apache/arrow-rs/blob/master/arrow/src/array/array_binary.rs#L846-L849

A straight way is to add the offset:

        let builder = ArrayData::builder(DataType::Decimal(precision, scale))
            .len(v.len())
            .add_buffer(v.data_ref().child_data()[0].buffers()[0].clone())
            .null_bit_buffer(v.data_ref().null_buffer().cloned())
            .offset(v.offset());

However, the answer is still not 100% correct when the offset of the child data > 0. (Nested offsets always make me confused 😅).

The way I prefer is to build decimal array from FixedSizeBinaryArray and drop this method, so that we can avoid nested offsets. Maybe we can file another issue to track this.

[tustvold]

This is related to #1726