Newtype fugit Rate, Instant and Duration

[bugadani]

Closes #2923

skip-changelog because of embassy/wifi updates.

[bjoernQ]

Thanks!

We could consider a From<core::time::Duration> impl but not necessarily in this PR (it's already big enough)

[MabezDev]

This LGTM, thanks for taking care of the impl @bugadani!

The only thing left to do now is bikeshed megahertz, I've offered my opinion, and a few alternatives.

[Ben-PH]

fugit forces the exclusive choice of u32 or u64. this creates frustrations when wanting to interface with other crates that use usize, or some other int-len.

Awhile ago, I ran into this issue, and it's a pain.

I would recommend something that bases itself more on using the typenum crate, so as to clear out this potential issue.

I've been doing some work on this recently, but it's unfinished and unrefined, but it certainly serves well as a PoC.

[MabezDev]

fugit forces the exclusive choice of u32 or u64. this creates frustrations when wanting to interface with other crates that use usize, or some other int-len.

This PR implements this plan: #2923 (comment) where we've chosen fixed backing types for our implementation, so no need to worry about that here :).

I would recommend something that bases itself more on using the typenum crate, so as to clear out this potential issue.

I'd like to avoid using typenum where possible, slow compile times and too many generics.

[bugadani]

Awhile ago, I ran into this issue, and it's a pain.

What's your use case?

[Ben-PH]

Use case was ages ago, so I forget the details, but the key issues is that there were incompatable types embedded into const generics, and just made the whole thing messy.

I'd like to avoid using typenum where possible, slow compile times and too many generics.

Noted. What do you plan to do for the cases where the types chosen for the const generics cause headaches?

[JurajSadel]

Thanks!

[MabezDev]

Noted. What do you plan to do for the cases where the types chosen for the const generics cause headaches?

In general, we're trying to avoid generics (even the const kind) in the public API, and instead to the type checking in the constructor then convert the driver to a "concrete" type.