SOAP is very slow on FLAMINGO high-res. high-z

[MatthieuSchaller]

Running the default script in the repo, the code times out after 72hours and seems to be very far from having completed.

[jchelly]

The changes in #64 might help with this. We're sorting the particles by radius more times than necessary and passing all of the particles in the largest aperture to every calculation. We should refactor it to do the sorting at a higher level in the call stack then we can identify the subset of particles corresponding to each aperture and only process those.

We could also do some profiling on subsets of halos to see if there's anything very slow inside the property calculations.

[jchelly]

The fixed physical apertures are a problem too. The amount of work to do per halo will scale as (1+z)^3.