better treatment for inflow BCs

[zingale]

For inflow BCs, we should adopt the strategy from Maestro and others where the ghost cell value represents the interface value, and not the cell-center. And then we should use modified stencils at the edge to ensure that we see the correct inflow state.

[maximumcats]

It would also be good to be more consistent about this throughout the code because we already treat the boundary values on phi as edge values.