checker,cgen: fix duplicates in generic sumtype

[MCausc78]

1. Fix cgen error for Name[int](123), when type Name[T] = string | int | T #19159
2. Add tests

PS D:\Games\Proekti\V\vl> type test.v
@[on_duplicate: 'thats sentinel lol'; notice_if_duplicate]
type Name[T] = string | int | T

fn main() {
        _ := Name[int](123)
}
PS D:\Games\Proekti\V\vl> .\v2 run test.v
test.v:5:7: notice: duplicate type: thats sentinel lol
    3 |
    4 | fn main() {
    5 |     _ := Name[int](123)
      |          ~~~~~~~~~~~~~~
    6 | }

Note: you can control, whether to raise a error, notice or ignore if user used type, which is present as T in generic sumtype

[spytheman]

hm, there is no existing precedent for using an attribute for toggling between an error and a notice 🤔. I am not sure if it is worth the complexity.
Would not either just a notice, or either just an error be simpler?

[MCausc78]

sometimes this is bad idea to give error. Consider that:
firstapi:

pub struct Row[T] {
pub mut:
	// children components
	components []Component[T]
}

@[on_duplicate: 'T is present in sumtype']
@[notice_if_duplicate]
pub type Component[T] = Row[Button]
	| Button
	| StringSelect
	| TextInput
	| T

secondapi:

struct Default[T] {
	v T
}

pub fn c[T]() Component[T] {
	return Component[T](Default[T]{}.v)
}

user (not dev of firstapi and secondapi packages), not knowing that e.g. TextInput is present in Component type, uses c[TextInput](), and gets an notice, instead of hard error. That also would allow deprecating duplicates in generic sumtypes.
Yet one example:

struct Undefined {}

@[on_duplicate: 'Sentinels are internal']
pub type UndefinedOr[T] = Undefined | T

That would give a error if user will do UndefinedOr[Undefined].

[spytheman]

@medvednikov what do you think?

[medvednikov]

I agree with @spytheman here. This complicates the language and gives too many options. I'd just stick with one option.

[MCausc78]

removed attribute and made error by default

[spytheman]

There are still 2 attributes:
ignore_generic_duplicates and on_generic_duplicate.

I do not know how to document/explain that in doc/docs.md.

[MCausc78]

Stale

[medvednikov]

Sorry, we'll figure something out to merge this.

[enghitalo]

I don't think using attributes or custom error messages is a good idea.