Pinpoint cause of chaotic divergence in multi-architecture comparisons

> That's expected. There's chaotic divergence as soon as some tiny floating point difference occurs anywhere. And for gun simulations we don't even simulate the same events so there's that too.

It would be really interesting to dig into this further, though. I don't think we can really get rid of differences in floating point behavior, but taking the concept of event seeds and applying it at a track level in Geant4 should be able to resolve a lot of these differences, I think.

_Originally posted by @wdconinc in https://github.com/eic/containers/issues/336#issuecomment-4895459931_

The issue names no files, tests, or entry points. Start by reproducing the multi-architecture gun simulations and comparing where floating-point differences and event sequences diverge. Investigate whether track-level event seeds in Geant4 reduce the divergence; completion criteria are not defined in the issue.

Written by the indexing model from the issue text.