Root functions should allow nullable args

[adamhathcock]

when doing .RootArg I noticed nullable values weren't being respected/propagated. This PR should allow that.

[adamhathcock]

I can retarget this PR when dev is made again

[NikolayPianikov]

Thanks for the PR.

Could you please describe the concrete scenario where a root argument needs to be passed as null and then injected into the object graph?

I intentionally avoided supporting this so far for a couple of reasons:

• Pure.DI tries to keep generated object graphs predictable and validated at compile time. Today, when a dependency is injected, the consumer can generally rely on the injected reference not being null; constructors do not need defensive null checks for DI-provided objects.

• Root arguments behave like normal bindings: they are values that can be injected anywhere in the root object graph. If we allow null for a root argument, the same value may later be injected into a constructor parameter/property that is declared as non-nullable. That weakens the current guarantee and may turn a compile-time wiring concern into a runtime nullability problem.

• Nullability annotations are not part of the runtime type identity. For dependency resolution, string and string? are still the same type. So allowing nullable root arguments is not just a signature-generation change; it affects how safely Pure.DI can propagate that value through the graph.

For optional data, I would currently expect users to model it explicitly, for example with a wrapper/options type, a domain-specific abstraction, or a default value, depending on the scenario.

I’m open to supporting this if there is a strong use case, but I would like to understand the intended pattern first and make sure it does not break Pure.DI’s expectation that injected dependencies are non-null unless explicitly modeled otherwise.

[adamhathcock]

Real quick: I'm making a factory class that takes a nullable string and doing something in the factory method.

Depending on the value stuff is done and a new class with some injected dependencies is made.

You're right that dependencies themselves shouldn't be nullable but passed args can be.

I can give more details and stuff on Tuesday. Typing on mobile now. Thanks!