Fix multilevel wdconvect about maestroex HOT 4 CLOSED

The next test to try is to prescribe a velocity field in each code (e.g., start with u=1, and then try more complex flows such as outward expansion) and see if the convective update to the scalars gives the same result. This should be easy to spot any differences.

from maestroex.

When the grids become adaptive, we need to do the FillPatching using perturbational quantities. Refer to original MAESTRO regrid.f90

from maestroex.

Note that the new temporal integrator (with constant dr) fails very early. If you look at the velocity in plotfile 0 it has gibberish in it. Probably not related to the AMR issue with the original algorithm but still needs to be fixed.

from maestroex.

This seems to be fixed with the implementation of AverageDown instead of Reflux (for the original algorithm). See issue #37

from maestroex.