add PartialOrd support to all structs

[TyOverby]

Lack of PartialOrd support was my only issue when porting some of my code over to Euclid.

It's especially useful to preserve determinism after producing shapes in a non-deterministic manner.
Example:

• Create a bunch shapes on the GPU (deterministically created; non-deterministic ordered)
• Sort them on the CPU using this PartialOrd implementation
• Now your collection of shapes is deterministically ordered!

Not the distinct lack of tests. I wanted to run this by folks here before spending too much time on writing tests.

---

This change is 

[SimonSapin]

This ordering is arbitrary and has no meaningful relationship to geometry. If all you need is determinism in the iteration order of a set that’s fine, but it seems wrong to me to implement “the” ordering trait for this. Other trait impls like Add, where they exist, are implemented to make geometrical sense.

Perhaps instead of sort you could use sort_by_key with some arbitrary deterministic key, perhaps based on Hash?

[kvark]

Looks like a duplicate of #234, where the disposition appears to be: reject (edit: actually, "wait").
TL;DR: it doesn't make math sense, and it's relatively easy to implement for your wrapper types (in the way you mean it! since there is no one way to do it).