Index space weather records by UTC day

[scottshambaugh]

This fixes a timing bug in space weather data, which was indexed by TAI rather than UTC. This meant the last 37 seconds of each UTC day read the next day's record. This changes results very slightly as a result (though this doesn't get picked up by the unit tests - you may want tighter bounds on the golden value checks).

[ssmichael1]

This is subtle. Good find. I'll merge over the weekend. Thank you.

[ssmichael1]

recommended alternative, pure integer math:

instant.rs:

/// UTC calendar day as an integer Modified Julian Day number.
/// Same leap-second convention as `as_mjd_utc`, but integer arithmetic only.
pub(crate) fn utc_day_number(&self) -> i64 {
    (self.raw - Self::MJD_EPOCH.raw - microleapseconds(self.raw)).div_euclid(86_400_000_000)
}

spaceweather.rs:

let day = tm.utc_day_number();
let first_day = first.date.utc_day_number();
if day >= first_day {
    let idx = (day - first_day) as usize;
    if idx < sw.len() && sw[idx].date.utc_day_number() == day {
        return Ok(sw[idx].clone());
    }
}

Let me know your thoughts.

[scottshambaugh]

I like that, amended and force pushed