PartialOrd for scalars

[Ilia-Kosenkov]

• Scalar<T> to initiate abstraction of common scalar properties
• PartialOrd for Rint, Rfloat, Rbool (NA-aware)
• min()/max() on Rint & Rfloat based on PartialOrd
• From<Rint> and From<i32> for Rfloat (note that i32::MIN_VALUE will not be treated as NA_real_)

Closes #570

[Ilia-Kosenkov]

@JosiahParry , ready for initial review :-)

[CGMossa]

LGTM, if it passes test.. then it should be merged.

[Ilia-Kosenkov]

Funny thing it does not really solve the original problem -- min() and max() require Ord , not PartialOrd. We cannot provide Ord of course, but at least now it should work like

vec![Rint::from(1), Rint::from(2)].iter().filter(|&x| x >= 2);

[yutannihilation]

That's what I was wondering about when I commented on #570. The reason f64 has min() and max() is not that f64 is Ord, but it implements its own min() and max().

https://doc.rust-lang.org/std/primitive.f64.html#method.min

[Ilia-Kosenkov]

Ah yeah, no problem! We already have some methods defined (mostly R-inspired), I can extend it to also include min(), max(), etc