Remove the default Send bound on traits

[alexcrichton]

See #10296 for the rationale, and commits for the implementation.

[flaper87]

LGTM. I always try to keep each commit "compilable" so, I'd prefer if these commits were squashed.

[huonw]

@flaper87 these cases are often a good time to break that (useful) rule, to make it super-clear what the core change is, and what the purely mechanical fix-ups are.

[flaper87]

@huonw agreed, I find them useful as well. Perhaps those could be squashed after the review and before the final approve. With that we'd loose the benefit of these commits being split in the history, though. Anyway, probably not a big deal.

[huonw]

There doesn't seem to be a way to specific bounds on generic trait objects (#9265). Neither of these parse:

traitFoo<T>{}let _: ~Foo:Freeze<int>;let _: ~Foo<int>:Freeze;

This PR is blocked on that issue.

[alexcrichton]

Why is this blocked on that issue? The issue is certainly more pressing, but it would be nice to get off the default bounds as soon as possible.

[huonw]

Because without the new parsing there's literally no way to have a sendable generic trait object; while the current situation just means you're forced to use sendable ones even when a non-sendable one is sufficient. (In any case, the point is moot since you opened #13079.)

[nikomatsakis]

👍 I'm happy to see this, I was just thinking "Oh, I should go implement that by now, might be something I can do in my spare time." Instead, I'll review. :)

[nikomatsakis]

We have got to change this printout. (not necessarily in this PR)

[nikomatsakis]

The changes look good to me. I say r+ pending resolution of @huonw's objection.

[alexcrichton]

@nikomatsakis, @huonw, with #13079 soon-to-land (hopefully!), are you guys ok with this?

@nikomatsakis, you may also be interested in #13155 as to why I had to write :Send so often when I was hoping to just write a supertrait bound on all the Rtio traits.

[flaper87]

@alexcrichton could you also update the docs? There's a line in the tutorial that says ~Trait:Send is equivalent to ~Trait

[alexcrichton]

@flaper87, good catch, I've updated the docs a bit.

[huonw]

r=me with minor docs edits