prefer_const_literals_to_create_immutables support for extension types

[pq]

Related to #59275,

extension type E(List<int> i) { }

var e = E([1]);

should we treat E as immutable and produce a prefer_const_literals_to_create_immutables lint on [1]?

/fyi @goderbauer @a14n @eernstg

[bwilkerson]

The documentation for the lint doesn't make it clear what the motivation for the lint is, and I don't remember it. Could you describe the motivation? It's hard to know how to answer the question without understanding what the lint is trying to accomplish.

[lrhn]

Presumably, the @immutable annotations is intended to convey that the class instances should be deeply immutable.
By passing in only deeply immutable arguments to the constructor, that is more likely to stay true, and the only way to ensure that arguments are deeply immutable is that they are constant expressions. (Or, presumably, instances of other @immutable-annotated classes.)

The documentation for @immutable isn't that clear. I'm guessing it shows intent to be deeply immutable, but the only direct effect it has on the annotated class is to require all fields to be final. This lint is trying to force you towards full immutability.

[pq]

Presumably, the @immutable annotations is intended to convey that the class instances should be deeply immutable.

I think this is right.

This lint is trying to force you towards full immutability.

And this.

[pq]

Fixed w/ 039640f.