Parameterized source distributions

[sethrj]

For testing it would be good to have a little more complexity in our RNG-determined input that would allow us to have a reasonably realistic parameterized set of "primaries" that could come from a Geant4 problem. To do this we should:

• Enable offload_output_file to ROOT and set up test beam inputs for HGCal, TileCal (and maybe ttbar inputs for full CMS?)
• Save a GB or so of primaries from each case
• Calculate some histograms for angle, particle type, energy, position (r/z?)
• ???
• Implement a new PrimaryGenerator that can sample from a relatively complex distribution
• Use that distribution to sample primaries in our regression tests