1. Open a text file in the IDE
2. Make sure you have some abbreviation defined for plain text (for example zzz -> expanded${cursor}text).
3. Go at the beginning at the file (the file doesn't need to be empty, just the position zero) and try to expand the
abbreviation (zzz+TAB or whatever your shortcut is).
4. You'll notice that while the text is expanded, the cursor remains on position zero. This isn't actually influenced if
"${cursor}" is present in text, the cursor will be misplaced every time.

The problem seems to be that we have a call similar to CodeTemplateManagerOperation.insert ->
CodeTemplateInsertHandler.processTemplate which adds in textSyncGroup some regions based on document Positions. The
problem with positions is that they get notified when the document changes.

So, later on, when you call formatter.reformat (in CodeTemplateInsertHandler.run) the positions will be notified and
will get reset to zero ! Then in TextRegionEditing.activateTextSync we call component.setCaretPosition with the zero
position.

The bug shows up in all the editor kits that use the old Formatter and the cause seems to be that
org.netbeans.editor.Formatter.reformat just seems to re-insert the same text for no reason. The side-effect of this is
that the Position is notified when the text was removed (and gets reset to zero) but not when it's re-added.

My patch just checks that the Formatter's writer actually changed something and only then re-adds the text.

Note: I'm not 100% sure it's a Position that gets added in TextRegionEditing as I didn't have the patience to go that
deep but I remembered the behavior from an older bug I tracked in the editor. The code in Formatter though does seem to
call remove/insertString for nothing when the text wasn't changed actually (which is bound to trigger some events from
the document for nothing).