Consistency in Parthenon Hook Function Arguments

[pdmullen]

PR #999 argued in favor of using MeshBlockData<Real> *mbd function arguments for a newly added PostInitializaiton hook. This inspired some discussion regarding consistency between hook function arguments.

For example, many hooks take MeshData<Real> or MeshBlockData<Real> arguments (which comes with the nicety that downstream users can template on typename to interface with both). However, the MeshBlock variation of ProblemGenerator is somewhat unique in that it takes *pmb (the Mesh variation takes as argument MeshData<Real>).

Here, I propose that we try to enforce some form of consistency between hook function arguments. In the example above, I'd advocate for requiring the MeshBlock variation of ProblemGenerators take MeshBlockData<Real> as function argument. Of course, this would be a breaking change... requiring downstream users to update all their pgen.