feat: crypto interface

[jacobheun]

This adds a crypto interface to js libp2p, we currently don't have one. This also includes a mock crypto file for tests, which demonstrates how a crypto module might be built, including a handshake and stream wrapping which can be used for crypto boxing/unboxing.

Aside from the clear change to async iterables, this also updates the "existing" interface to use the protocol property instead of tag to be clearer about its purpose.

I added the iterable wrapper logic to the js-libp2p docs in a PR, in case it's useful to others libp2p/js-libp2p#466

[mkg20001]

Is plaintext not going to be a module?

Edit: It would be useful to have it as one, so people who want to experiment with it can simply add it like "just another crypto module"

[jacobheun]

Yes, that's the intent, I'm planning on writing that and the private network refactor this week. I was planning on just bundling plaintext 2 with libp2p though, and linking to it here.

[mkg20001]

Ok, perfect. I'll keep this discussion open until that's resolved for tracking.

[jacobheun]

I also created #3 to track adding plaintext.

[jacobheun]

I've updated the docs and added a test for the expected remotePeer on an inbound connection. I also added an errors export with an UnexpectedPeerError as an initial value. We can add to this as we determine other, common errors crypto modules should throw.

[vasco-santos]

LGTM!

Just found a small typo

[mkg20001]

LGTM

[mkg20001]

I assume that with js-interfaces now being a thing, interface-connection is no longer needed and the functions return "just streams"
Is that correct?

[jacobheun]

I assume that with js-interfaces now being a thing, interface-connection is no longer needed and the functions return "just streams"
Is that correct?

Yes, that's correct. Now when you dial in libp2p the Connection object is not created until just before the dial returns. It's now basically just a metadata/convenience wrapper around the muxer and transport connection (MultiaddrConnection).