We need to avoid touching the age from a mark word that has a displaced header.

This is because `oopDesc::age` is subtly racy: we read the mark, check it for displaced header, and then reach for the displaced header, reading the mark again. If stack unlock happened at the same time, we might be reading non-displaced mark the second time, which would fail the assert or even reach `fatal` in `markWord::displaced_mark_helper`. (We should really be reading the mark once there, see JDK-8309953)

But it would be even easier to rectify the filtering in Shenandoah dedup to avoid going into oopDesc::age to begin with. This is easy, since we load the mark anyway.