Use the rust-version field in Cargo.toml to set minimum Rust version

[bradjc]

This encodes the minimum version so that the tooling will verify the user has a new enough Rust.

Set based on the needs from #498.

Fixes #394. Supersedes #499.

[lschuermann]

How does this relate to rust-toolchain? Would we eventually want to switch away from this entirely in favor of a MSRV?

Changes addressed

[lschuermann]

Seems like this needs #481.

[bradjc]

Ok yeah this applies on top of https://rust-lang.github.io/rustup/overrides.html

This is confusing, because either rust-toolchain meets this requirement or nothing will build. But, since we also use stable for tests, this is useful for making sure the tests compile with a new enough rust.

[bradjc]

Why don't we use stable by default and nightly only for miri?

[jrvanwhy]

Why don't we use stable by default and nightly only for miri?

I first introduced the check (#315) when only a subset of libtock-rs' crates built on stable, then later expanded the set (#390) to all crates. At the time, I had a few ideas for things other than Miri that required nightly, but I haven't had the time to implement them so at the moment we only need nightly for Miri.

I think it's fine to flip that around, and use the MSRV by default (e.g. via rust-toolchain and invoke our nightly version only for specific commands (at the moment, the Miri invocations).

[jrvanwhy]

This should be set workspace-wide and inherited.

... which requires a newer nightly toolchain, because the current toolchain is too old to support workspace inheritance.

[lschuermann]

Seems like someone should really be looking into updating the nightly toolchain, eh? 😆
I can attempt that tonight, if adding sptr as an external dependency is uncontroversial enough. All of the other proposals in #481 (especially mine!) seem ... less than suboptimal.

[jrvanwhy]

We need to test with this MSRV in Makefile rather than using +stable, otherwise this will bitrot.

[bradjc]

Done in #513