Related to JDK-8035557, one way to temporarily fix the issue with the wrong predictions of rs length and card prediction is to suppress sampling of these metrics during mixed gc.

Since mixed gc does not use these, sampling during mixed gc only inflates these predictions for the young gc following mixed gcs.

The disadvantage is that the prediction for this first young gc after the mixed gc might be off due to many leftover cards from the mixed phase.