Fused Time Stepping for LimitingADERDGSolver crashes with MPI if recomputation is triggered

Nonfused variants seem to work fine.

EDIT: A run with generic kernels and noarch did not crash.