Updates to instant and its methods

[nekevss]

This PR updates Instant to some of the newer specification changes.

TLDR of the below:

• Updates Instant struct
• Updates diffing ops and adds tests
• Implements FromStr.

Primarily, it brings since and until up to date with both temporal_rs and the specification as a whole (along with at least 2 basic tests for since and until)

This PR also changes Instant inner nanoseconds from BigInt to i128. My thought is to eventually remove all usage of BigInt from temporal_rs in general. Primarily motivated by the thought that if NormalizedTimeDuration is using a i128, I don't see a huge reason why Instant should differ, unless I'm overlooking something and someone thinks otherwise.

Finally, it implements FromStr for Instant. This will allow parsing of JsString in Boa, and would unblock to_temporal_instant.

[jedel1043]

Hmm, is this asserting an invariant? Because temporal_unwrap should only be used to ensure invariants are preserved. If this is used to unwrap errors caused by the user, then I think it should return an error instead, which is a bit more explanatory.

[nekevss]

It should be an invariant at this point. parse_instant requires a string with a DateRecord, TimeRecord, and OffsetRecord, and throws an error if any of them are None, which is what the NOTE was trying to highlight. 😆 Although reading it back now, the note probably makes it more ambiguous.

We can return an error in these places again, but they would be redundant errors that realistically would never throw.

[jedel1043]

Oh, now I see why. On parse_instant, we validate that the fields are Some and throw otherwise. Can we move that invariant to the typesystem? Instead of returning an IxdtfParseRecord, we return IxdtfParseInstantRecord with the validated fields being not Option.

[jedel1043]

Looks great!