HSE boundary conditions don't work with fv4

[zingale]

Not sure why, but they seem to cause the Rayleigh-Taylor to life up as time evolves.

[zingale]

note that it works just fine with compressible_rk using RK4 integration, so the issue seems to be the spatial reconstruction, not the temporal evolution. I suspect that the problem is that the 4th order stencils don't work because we are passing in cell-averages and then we need to do HSE integration on the cell averages, but we are treating them like cell-centers.

[zingale]

it would be better to just add the one-sided stencils to do a better job with reflecting BCs with 4th order, see #244

[zingale]

[zingale]

that's an example ^ left is HSE BCs and right is reflect