new ASIO host implementation

[LastExceed]

Closes #1299
Obsoletes #1083

[edwloef]

The non-azo dependencies seem unnecessary except for maybe oneshot, how much would it hurt to get rid of them?

• x.unique() -> { let a = x.collect::<Vec<_>>(); let mut b = a.iter().cloned().collect::<HashSet<_>>(); a.into_iter().filter(|x| b.remove(x)) }
• SmallVec doesn't seem to be on a hot path, or a path that needs to be realtime-safe, so it can just be a normal Vec maybe?
• x.pipe(y) -> y(x), x.pipe_ref(y) -> y(&x)

[edwloef]

except for maybe oneshot

and closure-ffi. That one in particular is absolutely indispensable.

Ah right, I had missed the cfg(windows) dependencies section. oneshot and tap should probable be moved down there as well, if I'm not mistaken?

[LastExceed]

I moved them to the platform dependencies for now to be consistent with what we currently have, but it might be worth revisiting this in the future as it could get rid of some duplicates. E.g. jack and audio_thread_priority are both specified for windows and linux with the same version, despite being optional in both cases

[Marco-Farruggio]

Any issue with channel count not being const fn? And can’t as u64 be used here? (Regarding the two MSRV versions)

[Marco-Farruggio]

I agree that there are definitely some places where rustfmt is in my opinion making code less readable, keeping a codebase (like cpal) legible and clear is important, but I think just being logical can exceed rustfmt in terms of neatness

[LastExceed]

Any issue with channel count not being const fn?

No, but I was hoping that we'd up the MSRV for .cast_unsigned() as mentioned earlier, in which case .channel_count() can stay const

can’t as u64 be used here?

where do you mean exactly?

[LastExceed]

The branch is finally fully ready to be reviewed and merged. Implementing this was quite the journey, I kept running into edge cases that required me to change my approach over and over. azo progressed from v0.0.9 all the way to v0.4.0 during all of this, and another project came forward to express satisfaction with the library (LastExceed/azo#6) in the meantime, which is very motivating.

I recognize that big PRs like this can be daunting to review, especially when everything is squashed everything into a single commit as I had initially done, so I spent the last 2 days rewriting the git history in order to make the PR as approachable as I can. Please let me know whether or not this turns out to be helpful.