remove abort_on_failure

[zingale]

At the moment, if abort_on_failure is set, we abort if the burn fails, but only on CPUs. If the burn is done on the GPU, then we don't. We should instead move this up to the application code, and make it responsible for checking the return code burn_t.success.

Castro already does this
MAESTROeX will once AMReX-Astro/MAESTROeX#379 is merged
I am not sure about Quokka (@psharda?)

[psharda]

Hi @zingale, yes, this is what we are doing in Quokka: we use burn_t.success to evaluate if integration in microphysics was successful, and this way it works both for CPUs and GPUs: https://github.com/quokka-astro/quokka/blob/development/src/Chemistry.hpp#L83