Update behavior around substituting Foo? into a usage of an unspec-bounded type variable

[cpovirk]

...if we decide in jspecify/jspecify#248 to change it: Currently, the result is Foo?, but we're discussing changing it to Foo*.

(I suppose someone could remind us that it wouldn't be wrong for a checker to continue to produce Foo?. However, presumably that's not what we want from the reference checker: At the very least, we'd really like to be able to cross-reference the reference checker's results against the conformance tests, which will include * where appropriate!)

I started writing a longer thing here about how the changes required might go beyond what the word "substituting" might suggest. However, I've gotten myself confused, and it's the end of the week, so I'll figure it out another day. It will end up fitting better into a discussion about jspecify/jspecify#248, anyway.)