Merge *SliceExt traits, use assoc types in SliceExt, Neg and Not

[japaric]

The following traits have been removed:

• ClonedSliceExt
• PartialEqSliceExt
• OrdSliceExt
• BoxedSliceExt

All their methods have been moved into the SliceExt trait using where clauses.

The Neg and Not traits now have an associated type for their return type. This breaks all existing implementations.

[breaking-change]

---

This requires a snapshot that contains PR #20374

r? @aturon
cc @nikomatsakis

[alexcrichton]

This looks amazing, thanks!

I think the only part about this we really need to think about is the name of the associated types (now that they're a public-facing API). The question of "what should I name this" will probably come up fairly frequently, so it would be good to have an answer. That being said, though, I'd rather land this now if we don't quite have an answer than take time to sort it out.

[aturon]

@japaric In case you didn't see on IRC: I just realized that using Result by default here is problematic because of the prelude, especially if we start bringing these into scope for the trait. Output seems like a good choice to avoid this problem.

I'm sorry for not catching this earlier on :(

However, I suspect that it should be easy to run a sed script over the patch itself to fix this. Feel free to squash into a single commit if that would make it easier.

[aturon]

OK, otherwise I've reviewed this PR as well and it looks fantastic.

[japaric]

Closing in favor of #20410