provide RandomMod::try_random_mod and Random::try_random methods

[baloo]

This method is to be used by downstream crates:

• bump rand_core from 0.6.4 to 0.9.1 RSA#478

Depends on:

• Relax Sized requirements for blanket impls rust-random/rand#1593

[baloo]

I think this is fine as random_mod is rejection sampling and it's not constant time in the first place, but please confirm that.

[newpavlov]

We need to decide whether we need explicit ?Sized bounds or not. If they are not needed, then we need to modify rsa accordingly. Otherwise, you can revert my two commits.

[baloo]

I think crypto-primes relies fairly heavily on the ?Sized

[fjarri]

So I started this whole ?Sized thing because it was mentioned in that RSA PR that switched to crypto-bigint. And also it seemed like we're not losing anything by relaxing this bound.

I have a use for it in another project where I need to pass an RNG through a dyn-safe trait API. Before it was available I just had a wrapper that boxed a dyn CryptoRngCore and could be used as a Sized argument, which was kind of inconvenient and added a minor performance overhead. That project also uses crypto-primes, so I relaxed the bounds there as well.

[newpavlov]

As discussed above, the ?Sized bounds are indeed necessary, so I reverted my two commits.

[Snowiiii]

When this is ready ?

[baloo]

I think it's mostly ready, but we might want to wait for a release of rand_core.

[newpavlov]

@baloo
BTW we generally use "squash and merge" for all PRs, so having a clear commit history inside PRs is not very important. Force push also breaks links on the "notifications" page, resulting in a "couldn’t find those commits" page, and makes it a bit harder to track changes during PR development.

(I know that some consider it a sloppy approach, but in my opinion it's more convinient and less time consuming for a forge-centric development)

[baloo]

This last push was fixing a merge conflict with #772
I've switched to using Jujutsu a month or so ago, and it's less natural to keep piling commits with it than it would be with a straight git. Sorry for the notification noise.