Fix: substitute generic type parameters when resolving typealiases

[shilei365]

Summary

When a generic typealias (e.g. typealias Foo<T> = Bar<T, Never>) is resolved, the placeholder type parameters from the definition were used directly in actualTypeName without substituting the concrete types from the usage site. This caused template code accessing variable.typeName.actualTypeName.generic.typeParameters to see the raw placeholder name (e.g. Value) instead of the actual concrete type (e.g. [String]?).

Fixes #1462

Reproduction

typealias MyPublisher<Value> = AnyPublisher<Value, Never>

protocol DataPublishing {
    var dataPublisher: MyPublisher<[String]?> { get }
}

Before: actualTypeName.generic.typeParameters[0].typeName = Value
After: actualTypeName.generic.typeParameters[0].typeName = [String]?

Difference from #1326

Issue #1326 (fixed by PR #1327) addressed method-level generic type parameters being incorrectly replaced by unrelated global typealiases. That fix added a guard to skip substitution when the type parameter name matches a method's own generic parameter.

This issue is about typealias-level generic parameter forward substitution — an orthogonal code path in actualTypeName(for:containingType:) that was never handled.

Changes

ParserResultsComposed.swift — In actualTypeName(for:containingType:):

1. Detect when the resolved typealias has generic parameters AND the usage site also has generic parameters
2. Build a positional mapping from RHS placeholder names to usage-site concrete types (filtering out parameters that are already known types like Never)
3. Substitute placeholders before constructing the returned TypeName

ComposerSpec.swift — 3 new test cases:

• Single placeholder substitution (MyPublisher<[String]?> → Result<[String]?, Never>)
• Multiple placeholder substitution (MyResult<String, Error> → Result<String, Error>)
• Concrete RHS types not substituted (Never remains Never when it exists in the type map)

Test Results

• 3 new test cases added to ComposerSpec.swift
• All 695 existing tests pass with 0 failures. No regressions.

[krzysztofzablocki]

thank you!