Add no-std support

[krnak]

Add support for no-std builds.

comments to this PR:

• no code changed, except src/error.rs where I had to avoid usage of thiserror (crate has not no-std support)
• module frost still depends on std
• crate still depends on alloc because usage of Vec in VartimeMultiscalarMul implementations
• tests/bincode.rs enabled only on std feature, because it depends on serde
• serde enabled if std enabled, because serde-conditioned compilation for module messages is missing #9

[krnak]

I am working on alloc-independent version.

[str4d]

utACK with cleanup suggestion.

[daira]

utACK

[krnak]

cleaned up & rebased . but not ready to merge

[SOLVED]
I need your comments on these two issues:

1. I would be a lot of work to make tests working in no-std (working with unstable no-std version of proptest, changing sources of randomness, etc.). May I just disable most of tests if std feature is off?

2. If I condition batch module by new alloc feature flag, whole crate could be alloc independent. Unfortunately I have no idea how to conditionally disable VartimeMultiscalarMul trait bound for Sealed::Point, when alloc flag is off. VartimeMultiscalarMul is used only for batching and its implementation depend on Vec. My proposition is to add naive no-std implementation like:

    fn optional_multiscalar_mul<I, J>(scalars: I, points: J) -> Option<ExtendedPoint> where ... {
        #[cfg(feature = "alloc")]
        {...}
        #[cfg(not(feature = "alloc"))]
        scalars
            .into_iter() 
            .map(|c| c.borrow().clone())
            .zip(points.into_iter())
            .try_fold(ExtendedPoint::identity(), |acc, (s, p)| Some(acc + p? * s))
    }

Is the aforementioned approach fine? Or do you see any better approach?

[krnak]

I return to this PR. It's ready for a review and merge. @str4d @daira

Issues from the previous comment:

1. I let tests to be std-dependent. I don't see any reason for testing in no-std environment. Furthermore this PR doesn't bring any new code, just compilation flags.
2. I found a simple way (code) how to condition VartimeMultiscalarMul trait by alloc feature. So the crate supports pure no-std builds now.

[daira]

utACK

[conradoplg]

Looks great, thanks! Just needs a small adjustment

[codecov-commenter]

Codecov Report

Merging #11 (3caf57b) into main (0e912de) will decrease coverage by 0.10%.
The diff coverage is 71.42%.

@@            Coverage Diff             @@
##             main      #11      +/-   ##
==========================================
- Coverage   86.75%   86.65%   -0.11%     
==========================================
  Files          15       15              
  Lines        1148     1154       +6     
==========================================
+ Hits          996     1000       +4     
- Misses        152      154       +2     

[conradoplg]

I've changed to 2021 edition to fix the resolver issue and solved conflicts.