[Property Wrappers] Fix a crash in merge-modules with wrapped parameters.

[hborla]

This is fallout from #36521 that caused a crash in merge modules when the source was compiled with -emit-module-source-info and -enable-testing.

After deserializing a property wrapper on a parameter, we check if the interface type of the parameter is valid. For property wrappers with init(projectedValue), this ends up going through PropertyWrapperBackingPropertyTypeRequest, which is called from ParamDecl::toFunctionParam. After the above PR, this request now depends on PropertyWrapperAuxiliaryVariablesRequest. The compiler ended up in infinite recursion (and eventually crashed) trying to synthesize the auxiliary variables for the wrapped parameter, because it got stuck trying to find the serialized location of the wrapped parameter. The fix is to return early from computing the backing property wrapper type in the case where the declaration came from a module file, because the auxiliary variables have already been computed and type checked anyway.

Note: I think a better way to handle this is to set the resolved type of the wrapper attribute in the constraint system (and get rid of the side cache for the backing type) so it’s always serialized with the resolved type, but doing this breaks a lot of interface printing and SourceKit tests, and I intend to cherry-pick this to release/5.5 so I'd like to keep the fix targeted for now.

Resolves: rdar://77804605

[hborla]

@swift-ci please smoke test

[hborla]

@swift-ci please test source compatibility

[hborla]

@swift-ci please smoke test macOS platform

[hborla]

@swift-ci please smoke test Linux platform

[hborla]

None of the debug source compatibility projects actually failed; looks like an infrastructure issue.