async: add DelayUs

[Dirbaio]

Add async DelayUs, mirroring the blocking one.

Unlike blocking, delay_ms has no default impl because it's not possible to have default impls in GAT-based async traits.

[rust-highfive]

r? @ryankurte

(rust-highfive has picked a reviewer for you, use r? to override)

[Dirbaio]

How do we handle the changelog check for the async crate? No need to changelog anything until first release, I guess.

[eldruin]

Seems fine to me, thanks!
Do you really need two different futures for the us and ms variants?

The changelog check does not seem to support more than one changelog file in a repository so I think we can just ignore it for e-h-a changes. It is not required for a bors merge, just slightly annoying for maintainers.

[Dirbaio]

Do you really need two different futures for the us and ms variants?

an impl may want to have a loop for the ms variant like in the blocking default impl, to avoid overflow. In that case they'd be different futures.

Also, for consistency I'd rather always do "1 method = 1 future associated type". This is what the future "async fn in traits" will desugar to. Also, users will almost never use the associated types, they'll always do foo.method().await, so having many of them is not bad for usability. This gives max flexibility to implementors with no downside.

[eldruin]

I see, thanks for the clarification.
Looks good to me!
bors r+

[bors]

Build succeeded:

• ci-linux (1.46.0, x86_64-unknown-linux-gnu)
• ci-linux (stable, thumbv6m-none-eabi)
• ci-linux (stable, thumbv7m-none-eabi)
• ci-linux (stable, x86_64-unknown-linux-gnu)
• ci-linux-test (1.46.0, x86_64-unknown-linux-gnu)
• ci-linux-test (stable)
• fmt