Fix guards OpenMP and GPU execution policy guards in signed_distance API

[kennyweiss]

A user reported an error in how we handle execution policy guards in the signed distance API. Specifically, we're using #ifdef in some cases but should actually be using #ifndef guards.

See https://github.com/LLNL/axom/pull/1191 for a proposed bugfix. (Specifically, the changes to signed_distance.cpp)

We might consider using this as an opportunity to clean up this section of the code. Specifically, should the SignedDistExec::OpenMP enum case be defined in configurations that don't have OpenMP? Or would it be better to avoid this problem by either

• defining only the valid enum cases for the configuration E.g. something like changing the enum definition from this https://github.com/LLNL/axom/blob/9ea7c93c4b802b4da444727b995e4f02df344b3a/src/axom/quest/interface/signed_distance.hpp#L87-L93 to

enum class SignedDistExec
{
  CPU = 0
#if defined(AXOM_USE_OPENMP) && defined(AXOM_USE_RAJA)
  , OpenMP = 1
#endif
#if defined(AXOM_USE_GPU) && defined(AXOM_USE_RAJA)
  , GPU = 2
#endif
};

• or, keeping the enum as it is and adding an isValid(SignedDistanceExec) method that can better isolate the checks

This might depend on expectations about how users are using the SignedDistanceExec enum.

[cyrush]

since this could be an issue in multiple places, should we move to use axom-wide execution polices? (instead of ones for specific algorithms)