Generator hintName collision

[czirok]

• I agree that my contribution may be licensed either under MIT or any version of LGPL license.

Include the namespace in generated source filenames to avoid hintName collisions when different namespaces contain types with the same short name.

The filename generation in TypeDataHelper is now namespace-aware while keeping nested type names and generic parameter counts as part of the generated name.

This prevents AddSource collisions in source generators that use TypeData.Filename, including the GObject subclass generator and the GTK template generator.

Tests were added for both GObject and GTK integration paths with identical Widget type names declared in different namespaces, verifying that source generation no longer fails with CS8785.

[badcel]

The code itself looks perfectly fine for me.

Unfortunately the tests seem not to work like intended. I'm fine with having a "stupid" test like mentioned in the comments to detect if "something" is broken. I this happens to often we can still think about moving tests for namespace verification in actual unit tests.

[czirok]

I reworked the test so it verifies the actual generator output.

It now creates two subclasses with the same class name in different namespaces, runs the generator, checks that no CS8785 diagnostic is produced, verifies that both namespace-qualified hint names are generated, and ensures that all generated hint names are unique.