add support for ed25519 authentication

[9072997]

Description

This adds support for connecting to servers that require ed25519 authentication.

Checklist

• Code compiles correctly
• Created tests which fail without the change (if possible)
• All tests passing
• Extended the README / documentation, if necessary
  • I don't think this is necessary, since it does not appear that the supported authentication methods are listed anywhere
• Added myself / the copyright holder to the AUTHORS file

Reference implementations:

• Java
• C
• C#

The ed25519 implementation was taken from here because it was portable C (and therefore easy to make work with cgo) and because it was very close to the code the reference C implementation was derived from.

The biggest issue with this is that it introduces a dependency on cgo. I spent a fair amount of time trying to get this going with the ed25519 implementation from the standard library (crypto/ed25519), but I lack the cryptography knowledge to figure out what is happening here on a high level so I can use the corresponding high-level functions from crypto/ed25519.

If cgo is a deal breaker (and there is not someone else on the team with the knowledge/time to implement this in pure go), would it be worth exploring an API to allow additional authentication methods as "plugins"? I could then implement this as a plugin and only users who needed ed25519 would have to worry about cgo.

[9072997]

I may have discovered an issue with this compiling under Linux. I will take a look at it this weekend.

[9072997]

this compiles under Linux now

[ldanielw1]

Thanks so much for working on this! Was wondering if there were any updates on this effort? Would be very interested in seeing this support added into go!

[methane]

I'm sorry, but I am very busy these days so I check only issue tracker.
Merging new features is postponed.

By quick looking, cgo is deal breaker. I won't merge this pull request. Please use @9072997's fork if you really need the ed25519 plugin.

[9072997]

Since cgo is a deal breaker I'm going to go ahead and close this PR. I'll open another if I ever manage to get this going in pure go.