signature: use Fn(&mut D) for *DigestSigner/*DigestVerifier

[daxpedda]

I honestly tried to find the right words for the documentation, but it all sounds rather off to me.

My suspicion is that we will actually need FnMut here if users intend to use some Reader API inside.

Companion PR: RustCrypto/signatures#1064.
Resolves #2003.

[tarcieri]

I think it would be good to add some comments about the expected contract of this function, namely that it should expect to perform all of its updates in a single invocation, and if called repeatedly it should perform the same sequence of updates

[daxpedda]

Update for the signatures repo is up: RustCrypto/signatures#1064.

[daxpedda]

FYI because Digest isn't a supertrait of Update I kinda did the implementation with FixedOutput + Update . But I could also write Digest + Update.

Also just noticing that if we actually do FixedOutput + Update we should also add HashMarker? Should I add HashMarker in the trait here?

EDIT: Went with Digest + Update for now.

[tarcieri]

@daxpedda this is kind of annoying re: Digest + Update:

error[E0034]: multiple applicable items in scope
   --> k256/src/schnorr/verifying.rs:119:53
    |
119 |                 msg.iter().for_each(|&slice| digest.update(slice));
    |                                                     ^^^^^^ multiple `update` found
    |
    = note: candidate #1 is defined in an impl of the trait `Digest` for the type `D`
    = note: candidate #2 is defined in an impl of the trait `Update` for the type `Sha256`

Maybe it should just be Default + Update + FixedOutput? (which AFAICT would also work for Shake256`)

[daxpedda]

Yeah, we can just use Default + FixedOutput + Update, but in implementations. I don't believe Shake supports FixedOutput, where we can just skip it.

I mentioned this before, but we might want to add HashMarker as well?

[tarcieri]

Aah, I see, it looks like it did in sha3 v0.10 but not in v0.11. Sure, HashMarker sounds good.