stdlib: Place @_semantics prespecialization requirement target under correct constraints

[AnthonyLatsis]

subscript.read requires Bound: Strideable, Bound.Stride: SignedInteger.

The substitution map refactor in #31895 revealed this with a crash. With a bit of guidance, I am happy to add a regression test for master too, if warranted.

[gottesmm]

@CodaFi I think @atrick understands prespecialization the best now (and how -Onone support works).

[AnthonyLatsis]

@atrick ping

[atrick]

This looks like a straightforward fix, so I approved it. I have to admit, I don't understand how the issue was exposed or what would break in the old version. The resulting symbol name should be the same either way.

[AnthonyLatsis]

@atrick I should have linked the explanation, sorry. The issue was exposed because substitution maps have become part of the identity of a BoundGenericType in #31895, resulting in Assertion failed: (ReInfo.getSpecializedType() == SpecializedF->getLoweredFunctionType() && "Previously specialized function does not match expected type."). The difference in substitution maps causing the assertion failure is due to subscript.read and the corresponding prespecialization point having different conditional requirements. It looks like the prespecialized read coroutine is currently receiving pattern substitutions with an invalid Strideable conformance. I don't know of any real-world implications to this though; perhaps the prespecialization is just left unused.

[AnthonyLatsis]

@swift-ci please smoke test

[atrick]

OK thanks. Makes sense. This PR looks like it's just adding some test flags though: #31893 [silgen] Add an extra swift-version 5 run to initializer tests

[AnthonyLatsis]

This PR looks like it's just adding some test flags though

My bad, fixed the link.

[AnthonyLatsis]

@swift-ci please smoke test