Allow async and sync nats to compile on MIPS and PowerPC

[protochron]

Now that nats.rs depends on a version of rustls that includes the latest version of ring, it can now be compiled to run on additional types of CPU architectures.

Specifically this converts all uses of AtomicU64 to use the type provided by the portable-atomic crate, which uses the native version of AtomicU64 on all architectures except MIPS and PowerPC where that type isn't supported.

I went ahead and updated sync nats with the latest version of rustls which inflated the size and scope of this PR beyond just swapping out the AtomicU64 implementation.

[protochron]

I just ran across https://github.com/taiki-e/portable-atomic which might be better because it would provide support for even more architectures
Went ahead and just used the portable-atomic crate instead.

[Jarema]

Thanks for this contribution!

I checked briefly the portable-atomics and it looks good, with minimal depedencies.
Will bench it soon and review it.
This is nice timing too, as we're now investigating options to support more architectures.

[paolobarbolini]

Going by the popularity I'd also consider crossbeam-utils crossbeam_utils::atomic::AtomicCell

[Jarema]

@protochron I would like this to be merged, however, it would be good to check the crossbeam_utils against the less know crate :).

[protochron]

@Jarema I think using crossbeam_utils directly would be a much more substantial change to the code. I'd have add some directives around the AtomicU64 imports and calls where we need AtomicU64s to using an AtomicCell instead which is effectively what the portable-atomics crate does already. Also the author of the portable-atomics crate is one of the owners of crossbeam-utils, so I'm pretty sure they've done a better job of implementing atomics on these platforms than I will :D

[thomastaylor312]

@Jarema I also have a vested interest in multiple platforms and I think I agree with @protochron here. FWIW, I don't think portable-atomics is less well known as it has 15 million downloads from crates.io and the maintainer is also a maintainer of crossbeam. Also, because it delegates to existing atomic implementations if they exist, it seems like a really strong pick to enable what is needed here

[Jarema]

Ok, point taken @thomastaylor312 and @protochron :). Those are really strong arguments.

Let's stick to portable atomics.
I will review it shortly.

@protochron can you please resolve the conflicts?

[Jarema]

@protochron I would like to merge this before doing a release, so a kind nudge to fix conflicts :).

[protochron]

Sorry about that! Just synced main and resolved the conflicts

[Jarema]

LGTM!