Two places in the code had a similar logic: computing a node's type from its dependencies. One of the branches actually involved a bit of duplicate logic: checking for the dependencies size first, but then Type.merge(...) also checks for the size again.

This PR moves that logic to a helper method: type_from_dependencies. We could simply call Type.merge(dependencies) instead, but for later refactors I have in mind overriding type_from_dependencies for some specific AST nodes. For example, an if node is currently bound to its then and else nodes, putting them in the dependencies array. But that's not needed: we could override type_from_dependencies for If to return the type union of the then and else node. That way we don't need to store an array of two elements in dependencies for every if in the program.

However, I'd like to do those optimizations in a separate PR to keep things small and easier to review.