Improve readability of code regarding OsmDataLayer.validationErrors and ValidatorTreePanel

[GerdP]

There is a very confusing relation between the field public OsmDataLayer.validationErrors and the private field ValidatorTreePanel.errors. In some situations both refer to the same instance of ArrayList, but this is not guaranteed by the code. I assume the reason for this are unit tests.
The content of OsmDataLayer.validationErrors is directly modified in ValidateUploadHook and ValidatorErrorExporter.exportData() reads it.
The content of OsmDataLayer.validationErrors is also modified when ValidatorTreePanel.errors refers to the same object. This is controlled in ValidatorDialog.
My current understanding is that we have a single instance of ValidatorDialog and ValidatorTreePanel but possibly multiple instances of OsmDataLayer, so it seems to be correct to keep the error list in OsmDataLayer.validationErrors.
Question is why we don't use that field in the validator classes?

[GerdP]

In 14857/josm:

see #17401, #17431

• make sure that list of errors is never null
• performance: avoid to sort list of errors when only elemets were removed from an already sorted list. This reduces time spent to rebuild the tree when primitives are removed or errors are fixed with the Fix button.

[GerdP]

I think the current code is a bit clearer.

Question is why we don't use that field in the validator classes?

It makes the unit tests less complicated.