App owned logins

[walkah]

This PR introduces dependency injection for initialization and account registration (along with the various username validation features). This allows apps to override these and potentially use alternative authentication backends / mechanisms.

It also introduces a "local" version that allows an app to init/register accounts in a self-contained manner like this:

import { LOCAL_IMPLEMENTATION } from "webnative/auth/local";
setDependencies(LOCAL_IMPLEMENTATION);
wn.initialise({ loadFileSystem: false })
  .then((result) => {
    const { authenticated, username } = result;
    if (authenticated) {
      console.log(`Logged in as: ${username}`)
    }
  })
  .catch((e) => console.error(e));

Closes #316
Closes #302

[matheus23]

Minor stuff only & a small question. Otherwise looking good to me :)

[matheus23]

We could switch out console.log with the debug logging functions, so they only print with setup.debug({ enabled: true })

[expede]

setup.debug({ enabled: true })

This is also generally togglable in the console directly — no need for managing it ourselves

[matheus23]

Just a Q for clarification: Why do we abstract these functions? We're using the same implementation for both local auth as well as auth-lobby based auth it seems.
Is this so that someone would in principle be able to implement their own "username database"? If so, is that possible today? I think we'd need email-like identifiers for that, i.e. something like matheus23@fission.codes usernames instead of matheus23. E.g. in the DNSLink fetching code.
Maybe it doesn't hurt having these abstractions in here today already, but I'm not sure. 🤔

[expede]

Yes, this is so that you can plug in any arbitrary lookup method, rather than forcing everyone to make it look like Fission's. For example, web3.storage has this in a big database table that's at a different endpoint.

[expede]

RE validation: services may have special validation rules for their service (e.g. no curse words in Spanish). I believe that the plan is to provide defaults and let users override them if needed.

something like matheus23@fission.codes usernames instead of matheus23

That's more of an interoperability question. Blaine calls this style the "Jesus of Nazareth" format, and I agree that it reads very well (in English at least)

[expede]

(But per service, you can imply the host, in the same way that Discord hides the nonce in your username unless there's a collision on that server)

[expede]

LGTM! Bump the version and...