Fix channel join regression and fix possibly joining parted channels

[xPaw]

This PR creates channels out of join string when creating a network to avoid reusing this join string on future reconnections.

This PR also fixes a regression which was introduced in 2.0-pre: #358.

[williamboman]

This could be simplified a bit;

const delay = 1000;
commands.forEach((text, idx) => {
    setTimeout(() => client.input({
          target: network.channels[0].id,
          text,
     }), delay * ++idx);
});

[xPaw]

Joins have to run after commands, so that won't work. I also didn't touch this code, only moved it to another file.

[astorije]

To be honest, I'm not sure the delay * ++idx statement is something I would call "simplified"...

[maxpoulin64]

Agreed with @astorije, plus apart from changing the functions to arrow functions (thus breaking older versions), it doesn't actually change much. idx*1000 would do the same, as forEach already provide the index and thus doesn't need to be incremented.

[maxpoulin64]

Looks good to me, 👍

[astorije]

What about:

channels = args.join.replace(/\,/g, " ").split(/\s+/g).map(function(chan) {
    return new Chan({
        name: chan,
    });
});

or, if you don't want that long first line:

var join = args.join.replace(/\,/g, " ").split(/\s+/g);
channels = join.map(...)

[astorije]

Nevermind, got ruled out in priv by @xPaw, who plans to use the channels variable in a further PR and cannot blindly overwrite it.

[astorije]

👍