Study memory usage of photon lists for object parallelization

[cwwalter]

We would like to possible ask GalSim to return a list of photons for each object for us and then build a photon list for the entire sensor which we could then pass off to the sensor model. This has several advantages:

• The photons would be “mixed up” like the real physical system and that is important for sensor effects like BF.

• Running in OpenMP for the sensor accumulation part in C++ we can effectively use threading.

• We can actually collect the initial photon list over the focal plane. So, with Josh’s code we can more naturally handle things like ghosts etc.

We need to understand how many photons are associated with a single sensor in different scenarios and what sort of memory usage this implies so we can compare with (for example) KNL memory. Then we can also consider how checkpointing should work in this scheme.

This is a tracking issue for reporting the photon requirements in imSim after getting the photon lists back from GalSim.

[cwwalter]

This issue is also covered in #244.