Generalize improvement of initialization type frames

[flo2702]

This fixes #640 and fixes #641 by generalizing the fix in #621 . While that fix worked by overriding the error reporting for (common) assignments and method calls in the Initialization Visitor, this fix is general and assigns the correct type to the receiver in the first place in the Initialization ATF, so that overriding specific error-reporting is now unnecessary.

This will obviously conflict with #623 , which uses the overridden error-reporting functions to modify the error messages, so we need to consider how to handle this. The simplest way is to have #623 override this branch, which has the slight disadvantage that some of the code between the Visitor and ATF will then be doubled. Also #623 has the same problem as #621 in that it only applies to (common) assignments and method calls, not, e.g., to the example from #640.

[wmdietl]

@flo2702 Thanks a lot for the quick fix of these two issues!
This new way looks much nicer.

I was thinking whether to remove reportCommonAssignmentError, because now there is no longer any override and only a single call. But I think the symmetry to reportMethodInvocabilityError is nice and that method is still overridden once.

I expanded the tests a bit and all looks good.
The change to Issue5174.out is just new store numbers this time. A bit mysterious, but fine.

Thanks!