Should we analyze tuples element-wise?

[jcouv]

I will collect a few examples to show how we treat tuples element-wise in type inference and with regards to variance, but we don't do so for nullability analysis.

#nullable enable
class Program
{
    static void F((object?, int?) t)
    {
        _ = ((object, int?))t; // not great warning CS8619: Nullability of reference types in value of type '(object?, int?)' doesn't match target type '(object, int?)'.
        _ = (object)t.Item1;

        _ = ((string?, int))t; // not great warning CS8619: Nullability of reference types in value of type '(object?, int?)' doesn't match target type '(string?, int)'.
        _ = (int)t.Item2;
    }
}