Should we expose task::AtomicWaker?

[yoshuawuyts]

As per async-rs/futures-timer#39 it seems there are some scenarios where having an AtomicWaker type available would be useful.

It's somewhat niche, but I was wondering if it would perhaps make sense to expose it? Thoughts?

If we do we should probably create a separate crate so we can share the impl with futures-timer. Thanks!

[skade]

Hm, I'm not quite sure if we should expose that or if we should maybe have a second library as a "futures development kit". Though futures-util is already that.

Note that a good chunk of async-std uses no I/O and no dependencies (or just really small dependencies). Perhaps we should think of factoring that out into async-core or putting everything else behind a feature flag (perhaps named std).