Michael Zingale

this support will help https://github.com/AMReX-Codes/amrex/pull/3633

I'm looking at this some more. It is not related to reactions -- disabling reactions seems to make no effect. Change to `grav_source_type = 4` does help in general, so...

I think that the cleanest way to do this is to create a function `add_geometry_sources()` that is called in `trace_ppm.cpp` when we are doing the reconstruction of the other source...

I think this will require renaming some test suite benchmarks

this mostly works, but does not seem better than RKC. Need to experiment some more.

* `restart_with_vel_field` is also never used.

closing this -- this was an experiment for visualization

this looks a bit clumsy, so comments welcomed. With this, I can make `BackwardEuler` and `RKC` reuse the same `integrator_setup`

I am not sure how to fix the CUDA issues here. It seems to not like passing the integrator function into the `integrator_setup` function

update: `std::function()` doesn't work on the device, but a simple function pointer does.