na::partial_cmp is incorrect / confusing

Was writing some simple code using `nalgebra = "0.32.2"` and came across cases where `partial_cmp` was returning `None`. Is this an actual bug or is there some underlying mathematical "gotcha" where the partial cmp doesn't exist?

```rust
extern crate nalgebra as na;

fn main() {
pub type Pt2 = na::Point2;

// Succeeds
assert!(na::partial_cmp(&Pt2::new(0.0, 0.0), &Pt2::new(0.0, 0.0)).is_some());
assert!(na::partial_cmp(&Pt2::new(0.0, 0.0), &Pt2::new(0.0, 1.0)).is_some());
assert!(na::partial_cmp(&Pt2::new(1.0, 0.0), &Pt2::new(0.0, 0.0)).is_some());

// Fails
assert!(na::partial_cmp(&Pt2::new(1.0, 0.0), &Pt2::new(0.0, 1.0)).is_some());
assert!(na::partial_cmp(&Pt2::new(1.0, 0.0), &Pt2::new(0.0, 2.0)).is_some());
}
```

Edit: looking at the source it seems like it's checking that all axes have the same relationship. This is probably something that should be documented.