Consider restricting `Stream.ofNullable` to non-null type arguments

https://github.com/jspecify/jdk/blob/3669c24f878cf1dbaf42ccd75b002cdbc94fd4da/src/java.base/share/classes/java/util/stream/Stream.java#L1471C57-L1471C67

We could change the declaration of T to just <T>. That would be "simpler," and it could maybe in theory help someone who isn't realizing that the output is a Stream<@NonNull E> for some parametric-nullness type parameter E? But it might also annoy people who do actually want a Stream<SomeTypeThatMightIncludeNull>.

(TODO: Dig up past discussions of https://github.com/google/guava/issues/1927#issuecomment-68305122 through the lens of nullness. The idea here would be that you can always map(t -> t) to change the type if you want to. [edit: Here we go: jspecify/jspecify#525. And I guess maybe this issue fits better in the jspecify/jspecify repo unless we do want to start using this repo's issues more?])

One thing to note is that the current signature works well for the Checker Framework for converting a Stream<@Nullable Foo> to a Stream<Foo> via flatMap(Stream::ofNullable). I imagine that the proposed signature would, too, but it would be good to check.

Start with the Stream.ofNullable declaration in src/java.base/share/classes/java/util/stream/Stream.java at the linked line. Review the linked Guava and JSpecify discussions, then check whether the proposed signature preserves the Checker Framework conversion described in the issue. Done means reaching a justified decision about the signature and documenting the nullness and compatibility implications.