Timestamp Interval Arithmetic Ignores Timezone

[tustvold]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.

The timestamp arithmetic added in #4038 fails to correctly account for timezones

#[test]
fn test_timezone_interval() {
    let tz_str = "-01:00";
    let tz: Tz = tz_str.parse().unwrap();

    let array =
        TimestampNanosecondArray::from(vec![2419200000000000]).with_timezone(tz_str);

    let value = array.value_as_datetime_with_tz(0, tz).unwrap();
    assert_eq!(value.to_rfc3339(), "1970-01-28T23:00:00-01:00");

    // Add 1 month
    let interval = IntervalYearMonthArray::from(vec![1]);
    let date = add_dyn(&array, &interval).unwrap();
    let output = date.as_primitive::<TimestampNanosecondType>();

    let value = output.value_as_datetime_with_tz(0, tz).unwrap();
    assert_eq!(value.to_rfc3339(), "1970-02-28T23:00:00-01:00");
}

The above test should pass but it instead fails with the result 1970-02-27T23:00:00-01:00. This is because the month addition is done in the UTC epoch instead of the epoch of the array's timezone.

Describe the solution you'd like

Describe alternatives you've considered

Additional context

[alexandreyc]

I started working on this one 🐛

[tustvold]

Thank you, let me know if you get stuck. I would recommend making use of https://docs.rs/arrow-array/latest/arrow_array/temporal_conversions/fn.as_datetime_with_timezone.html and punting the non-trivial timezone logic onto chrono. In particular correctly handling DST is complex

[tustvold]

label_issue.py automatically added labels {'arrow'} from #4546