Replace spin with critical-section for no_std sync

[loadingalias]

I'm thinking about replacing the spin dep w/ critical-section for no_std synchronization needs.

While I was working on the no_std build fix (#34), I noticed that spin requires platform atomics that aren't available on all embedded targets (e.g., Cortex-M0). The current fix makes spin target-conditional, which works, but I wanted to float a potentially more robust long-term approach to no_std/wasm usage.

My idea is to basically replace spin with critical-section based sync primitives. This is the standard approach in the Rust embedded ecosystem and would:

• Support a wider range of no_std targets
• Allow users to provide platform-specific impls via the critical-section crate
• It also aligns w/ embedded Rust conventions - at least, I think it does? I don't have a ton of no_std experience. Maybe someone else could chime in?

It wouldn't come w/o a few trade-offs. We'd have the extra dep critical-section when I feel like we're trying to move in the opposite direction? Maybe I'm wrong here, but I'm under the impression we're aiming to reduce the deps. We'd also have the slight API differences from spin but it would ship as a major bump anyway.

I'm not suggesting this NEEDS to happen; I guess I just wanted to share the thought? I feel like we could do better w/ respect to no_std... and kind of think beyond being 'compatible' and more to 'awesome to use'. IDK.

Thoughts @onethumb?