derived variables are not high-order

[zingale]

The derived variables in the plotfiles are only computed to second-order accuracy. We may want to handle this via AMReX directly.

[zingale]

I think the way to this is to just fill ghost cells on the level for the derive MF, convert to centers, do the derivation including 1 ghost cell, and then convert back. But... we do this using a mask at C-F interfaces so we don't interact there. This means using 1-sided Laplacians at the C-F interface for the transformation from centers to average and back (like we currently do for the physical boundaries).

[zingale]

So in order to implement this, we need a mask.