cgen: fix codegen for option on concatexpr

[felipensp]

Fix #22951

_{Huly®: V_0.6-21405}

[Delta456]

This shouldn't be working as the language already disallows !?type. We also discussed this in discord.

[spytheman]

This shouldn't be working as the language already disallows !?type. We also discussed this in discord.

This is not !?Type however.

It is a result tuple, that just happens to contain an option type. Option type values should have first class support in V, and so should be able to be used everywhere other primitive types are allowed, including in result tuples.

[spytheman]

The problem with !?Type, is that it has 3 possible kinds of values, while the language has constructs like or blocks and if x := f() {, that can deal well with 2 at a time, and it is very confusing to use them, when there are 3 - one of those 3 will get forgotten, or cause bugs.

The only construct that can deal with 3 at the same time, is match x {, but that is very verbose to be used in all places that universal support of the old ?!Type combination would need.

For result tuples/multiple result values, with options inside them, there are no such problems, because the outer result, will get handled right away in one construct, and then what remains will be a normal option value, that can then be handled with the usual V constructs:

fn f() !(int, ?int) { return 1,2 }

a,b := f()!
// a will be `int(1)` and b will be `?int(2)`, and you can do `b or { ... }` etc

[spytheman]

Excellent work.