Halo Refinement Threshold
Prevent that the mesh refinement iterations are triggered to often.
During the mesh refinement iterations, we refine the Leaf cell parents of those Virtual cells that carry a refinement status greater than zero.
Currently, the time stepping iterations trigger refinement as soon as a Virtual cell carries positive refinement status. Especially with low order ADER-DG plus Limiter, mesh refinement is triggered quite often, which seems to be related to the CFL time step size: "Shock does not stay "long" in the same cell".
During the time stepping iterations, allow Virtual cells to carry a refinement status smaller some positive threshold value. Only if this threshold value is reached, trigger that the mesh is adapted.
Additional, we could think about allowing erasing only every $k$ mesh adaptions (=set of mesh refinement iterations).