fix: use core instead of std

[xtqqczze]

No description provided.

[clarfonthey]

Mostly in favour of this change, but I would rather not update the doctests specifically since in general we prefer making those as simple as possible, and using only std allows better import grouping.

[xtqqczze]

Import grouping shouldn’t be an issue here. In the doctests, we only use the following std imports, and they are never used in combination with core:

/// use std::hash::RandomState;
/// use std::rc::Rc;

[clarfonthey]

Would brace-group these:

Suggested change

	use std::collections::BTreeSet;
	use std::panic::catch_unwind;
	use std::sync::Mutex;
	use std::{
	collections::BTreeSet,
	panic::catch_unwind,
	sync::Mutex,
	};

[xtqqczze]

I had considered that, but rustfmt wanted to remove whitespace which I didn't like:

use std::{collections::BTreeSet, panic::catch_unwind, sync::Mutex};

[clarfonthey]

I see; personally, I still would go with it, but it's up to you.

[xtqqczze]

In an ideal world, we'd use imports_layout = "Vertical" like libc but it's an unstable feature.

[xtqqczze]

@clarfonthey By the way, would you prefer a fixup commit or rebase?

[clarfonthey]

For the main repo we don't want nested merge commits so a rebase would be ideal. Don't really mind if the result is multiple commits, although figured you'd also squash them as well anyway.

[clarfonthey]

If you want to propose new rustfmt.toml changes, wouldn't actually mind that, since formatting in CI is always via nightly. But probably better as a separate PR which can be ignored on blame.

[xtqqczze]

Right, yeah. Since the repo doesn’t use squash merges, I think a rebase is needed despite any reviewer preference.

[clarfonthey]

FWIW: there are clippy lints for alloc_instead_of_core, std_instead_of_core, and std_instead_of_alloc. Don't think they can be set up to exclude doctests though.

[clarfonthey]

At the moment, LGTM minus the clippy lints to help avoid this in the future.