`simple_convection` problem encounters negative density with SDC 4th order

[zingale]

When running simple_convection built as:

make USE_TRUE_SDC=TRUE
./Castro2d.gnu.MPI.TRUESDC.ex inputs_2d castro.time_integration_method=2 castro.sdc_order=4

we get negative densities at the top of the atmosphere. This is likely because of the average -> cell-center conversion. We should adopt the strategy from pyro where we mask a zone and drop to second-order if we violate rho < 0 or e < 0 (see
python-hydro/pyro2#309)