Different invalid SSAValue for unexpected initial_incoming_vals

[oxinabox]

There are two conditions for which we use negative SSAValues,
to trigger errors later if those values get used.

I figure we might as well make those different values,
because that will make debugging anything that goes wrong easier.

I say this after thinking I was debugging one kind of error, but I was actually debugging another.

[Keno]

Seems sensible. In the fullness of time we should have the verifier catch these things.

[vchuravy]

Should have some kind of test + documentation

[Keno]

We don't generally do either for internal assertions, which is essentially what this is.