checker: reject unresolved types used in interface match/is branches

[rilaaax]

Fixes #26972

Problem

Matching on an interface value with a branch type that was never actually declared (e.g. a typo, or a leftover/renamed type) compiles without any V error, but crashes the underlying C compiler with a confusing error instead:

cc: ...: error: unknown type name 'main__Nil'

Minimal repro:

interface Value {}

fn (v Value) get_str() string {
    return match v {
        Nil { '' }      // `Nil` is never declared anywhere
        else { '' }
    }
}

Cause

The parser registers an unresolved type name like Nil as a .placeholder type symbol (the normal mechanism used to support forward references). The checker is supposed to reject any .placeholder that's still unresolved by the time it's actually used, but type_implements_with_mut_receiver() in checker.v never checked for this. It validates a type against an interface by iterating over the interface's required methods/fields; for an interface with few or none of those (e.g. an empty interface Value {}), the loop does nothing and the function returns true by default — regardless of whether the type it was given actually exists. The bogus placeholder type then flows unchecked into cgen, which generates casts and variant-index machinery for a struct that was never defined, and only cc catches it.

Fix

• checker.v: type_implements_with_mut_receiver() now returns false immediately when given a .placeholder type, without erroring itself (some callers already have their own dedicated handling for this case).
• match.v: the interface branch of match_exprs() now reports unknown type '<name>' when the branch type doesn't implement the interface and is unresolved, mirroring the existing sum type variant error.
• infix.v: the is/as interface check is skipped when the type is a placeholder, since that case is already reported a few lines above, avoiding a duplicate error message.

Test

Regression test added.

[medvednikov]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 207abfa224

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[GGRei]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: d85f921627

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[GGRei]

The direct unresolved type case is fixed. To address the second Codex review, the checker still needs to reject unresolved placeholders inside generic instantiations, such as Box[Missing], with a regression test covering that case.

Thanks!

[GGRei]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c80b3b2c37

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[GGRei]

One Codex point is still only partially fixed. I reproduced it on a clean checkout: Box[fn (Missing) int] still reaches C generation and fails with an unknown type error. Please extend the unresolved-type traversal to ast.FnType parameters and return types, and add a regression test before closing “Recurse into container type arguments.”

Thanks!