centralized HSE boundary about castro

[zingale]

we want this to be handled automatically. All of the *fill routines in Castro first call filcc which handles the cases of EXT_DIR, FOEXTRAP, HOEXTRAP, REFLECT_EVEN, and REFLECT_ODD.

EXT_DIR falls through -- this has to be handled by the user. We can hook in the hydrostatic routine right after the filcc routine, so it is handled automatically.

[zingale]

BoxLib/Src/C_BaseLib/BC_TYPES.H says that EXT_DIR is data specified on EDGE (FACE) of bndry

but this doesn't seem to be how we use it -- we fill it as cell-centered.

[zingale]

There are a lot of implementations in the problem setups. Here's a summary:

• full integration, general EOS
  • xrb_mixed : this seems to be the most thorough -- we should base off of this
  • flame_wave : this is almost the same as xrb_mixed, but interpolates T from a model, different sign on momentum for zero-gradient
  • toy_convect : basically the same as xrb_mixed
  • test_convect : synced up with xrb_mixed
  • convective_flame : seems to do something compatible with xrb_mixed, but has it isolated in a separate function from bc_fill already
• analytic function or interpolation:
  • gamma_law_bubble : this just computes the analytic model
  • reacting_bubble : this just interpolates from the initial model using model_parser
  • double_bubble : does interpolation from some custom function

[zingale]

so it seems like we want 2 different BC fill routines -- one based on xrb_mixed that does HSE integration directly, and one that uses model_parser and interpolates

[zingale]

it looks like these all rely on constant gravity, but that's the only place I am aware of that we've needed HSE BCs. But we should probably put a test for this.

[zingale]

we also need to automagically handle the gravity, reaction, and rotation fills for these pre-defined EXT BCs

[zingale]

flame_wave was converted over to the centralized BC and the diffs look clean, but the test suite shows moderate errors

[zingale]

convective_flame handles this properly

[zingale]

see also #154

[zingale]

all of the problems that cane be, aside from test_convect, are now centralized

[zingale]

okay.... all normal HSE BCs use the centralized version now.