Add async functionality

[danielstuart14]

This new feature is gated behind the async feature flag, with the default behavior and syntax maintained.

It allows for asynchronously waiting for a pool slot to become available, making it easier to share constrained resources between multiple tasks.

It requires the AtomicWaker functionality from the embassy-sync crate, which in turn requires a critical section implementation.

Check the example for more details.

[danielstuart14]

I had to remove the change to portable-atomic as it was failing the CI/CD. I'll investigate it later and create a separate PR.

Let me know if this is something desired, as well as if the code is as you would like.
Thanks!

[Dirbaio]

I think this should be a separate struct. Having an async Cargo feature means you either get async support on all pools or no pool. If you only need async support on some pools you're wasting RAM for wakers that won't be used in the ones where you don't use async.

Also in general I'm a bit skeptical of doing "wait until there's free memory", it makes it too easy to get deadlocks if for example the occupied memory is owned by the current task.

Also, I consider this crate "done", I'd prefer to not add more functionality. I'd suggest you work on this in a separate crate.

[danielstuart14]

I think this should be a separate struct. Having an async Cargo feature means you either get async support on all pools or no pool. If you only need async support on some pools you're wasting RAM for wakers that won't be used in the ones where you don't use async.

Also in general I'm a bit skeptical of doing "wait until there's free memory", it makes it too easy to get deadlocks if for example the occupied memory is owned by the current task.

Also, I consider this crate "done", I'd prefer to not add more functionality. I'd suggest you work on this in a separate crate.

That makes sense, I'll move this to a separate crate! Thanks!
I agree with you that blocking until memory is available isn't good, but with async, we can use a timeout and join it to be sure it won't deadlock, that's the reason I developed it.

Thanks!