remove prefer_void_to_null from recommended

[pq]

In dart-lang/sdk#59309, @eernstg proposes we eliminate prefer_void_to_null and I'm inclined to agree. Unless there are strong reservations here, I'd propose that as a change to make before publishing 3.0.

/cc @devoncarew @lrhn @natebosch @jakemac53 @goderbauer

[devoncarew]

For other readers, the docs for the lint are here:

https://dart.dev/tools/linter-rules/prefer_void_to_null

And in-lined from the dart-lang/linter discussion, from @srawlins:

I agree I think it has outlived its usefulness; I was getting analysis_server compliant and it was pretty awkward; there were plenty of valid uses of Null.

Which sounds like the lint may be producing more false positives than we'd want.

[jakemac53]

It is worth noting that in particular for return types, void can be annoying, because it can interact poorly with LUB and other similar computations:

var x = switch(1) {
  1 => true,
  _ => print('weird!'), // because this returns void, `x` is now inferred as `void`.
};
print(x); // Error, can't use void

This is actually a simplification of a real situation that came up recently 🤷

[goderbauer]

I am not familiar with all the details here, but I have no objection to removing this lint. While it hasn't caused any trouble in our code base, it also hasn't been all that useful.

[lrhn]

I agree with removing this. As discussed elsewhere, there are cases where you should prefer void to Null, but with null safety, there are more cases where you'd actually choose Null deliberately, and you're less likely to choose the wrong one by accident.
Also null is no longer so special that it deserves the extra treatment. The same argument applies to every singleton type, whether () or a one-element enum (why pass a value to look at, if you know ahead of time what the value will be, it's better to say void so that you're not encuraged to actually look at the value.

The problem that this lint solves is now small enough, and well known enough, that it can be handled in code-review instead.

[natebosch]

LGTM. I have slight reservations that the same patterns this lint was intended to avoid might start coming back, but I do think it's less useful with null safety, and I do think I recall some places where I wanted to violate this lint.

[devoncarew]

This proposal is accepted.