feat: prevent wildcard aliasing

[MarinPostma]

This pr prevents wildcard aliasing. This is done in the ensure_column_name method, by checking if the considered column is a wildcard, in which case, we set its name to "*".

This is pretty hacky, by also works with ctes.

Maybe a better approach would be to change ColumnDef to something like:

enum ColumnDef {
    Wildcard,
    Def {
        id: CId,
        name: Option<String>,
        expr: Expr,
    }
}

By my knowledge of the project is not yet good enough to judge all the implications.

I haven't re-enabled any tests yet, it makes more sense to do that once #1094 is merged.

[max-sixty]

(I've seen this but I think @aljazerzen by far the better reviewer — IIUC he is traveling but is back quite soon)

[max-sixty]

Maybe a better approach would be to change ColumnDef to something like:

I agree, this is also similar to the proposal for now name resolution would work; now outlined in the internals in the book. But yes let's see what @aljazerzen thinks

[aljazerzen]

I'll do a refactor and add more info on how to do this.

[aljazerzen]

This is clever but a bit hacky. Passing a "*" does indeed prevent alias to be emitted, because it uses rule that SQL assumes alias from ident names only, so we omit explicit alias.

If we don't want to rely on this rule, I have to do a little refactor.

[aljazerzen]

Hey, my little refactor got out of hand :D I ended up doing exactly what you suggested @MarinPostma . I wanted to give more directions on how to do it, but it ended up being very connected to many different parts of the compiler, which is not ideal for onboarding.

This is the commit that partially does this:
4aae055

And this is the other part:
fd5b5c4

[max-sixty]

@MarinPostma thanks for the effort — we're v happy to have you as a regular contributor, and there's much more to do!