fix #210116: Delete all and Undo duplicates the clef

[mattmcclinch]

See 210116: Delete all and Undo duplicates the clef.

[anatoly-os]

Personally, I like it. @wschweer could you please review?

[mattmcclinch]

I replaced more occurrences of Measure::add() and Segment::add() with Score::undoAddElement(). I could produce the same error after trying to delete a courtesy key signature, and I got warnings after trying to delete bar lines.

[wschweer]

It should not be possible to delete a generated element. For the system header the clef segment should not be deleted if empty but set disabled. It will be reenabled later by layout when generating a new clef. So no undo/redo record is necessary for the clef segment. This also avoids confusion on undo.
I have another patch which tries to address this and at least fixes the clef problem.

[Jojo-Schmitz]

So this pr could be closed, right?

[mattmcclinch]

@wschweer, thank you for your comments. I have tested your solution, and I would like to make the following notes.

1. Since it is possible for an element to lose its “generated” status, it is still possible to delete an element that was originally generated automatically, by first making it invisible (for example). Perhaps it would be helpful to add a “modified” flag, and keep generated elements “generated”. That way, it would be impossible to delete a required element, even if it was modified after being generated.

2. In Score::undoRemoveElement() in libmscore/edit.cpp, when deciding whether the segment should be disabled or removed, perhaps a better test would be
   if (s->header() || s->trailer())
   instead of
   if (s->isHeaderClefType()).
   I can still cause a crash related to this issue, but this change seems to fix it.

[wschweer]

I changed undoRemoveElement() as you suggested.