Shared secondary bank performance optimizations

[paulromano]

Description

When doing profiling on a coupled neutron-photon transport simulation with weight windows enabled (which turns on the new shared secondary bank by default), I was noticing a lot of time spent in areas I wasn't expecting. This PR is the first round of some performance optimizations to address the issues:

1. First, I was noticing that a lot of time was spent allocating memory for the neutron/photon cross section caches (neutron_xs_ and photon_xs_ in the ParticleData class). At first, I made some attempts to only allocate the cache if it was needed for a given particle. Later on though, I realized the core issue was that in the main loop over the shared secondary "read" bank, we allocate a Particle on the stack. Thus, the cost of the memory allocations can be completely eliminated by simply allocating the Particle once per thread and re-using it in the loop (just requires clearing the local secondary bank at the end of each iteration on the inner loop).
2. I was also noticing a lot of time being spent in event_death. There are some atomics in there for global tallies that we can avoid in a lot of situations, so I've simply added conditionals to avoid triggering the atomics unless they are really needed.
3. The logic for RNG seed skip-ahead was being repeated for every RNG stream even though it's the same in each case. I've factored it out into a future_seed_coefficients function so that when we call init_particle_seeds, we only pay the cost once rather than 4 times.

I have another round of optimizations having to do with the shared secondary bank data structures directly that builds on top of these, so once this PR is completed, I'll submit a follow-on one.

Checklist

• I have performed a self-review of my own code
• I have run clang-format (version 18) on any C++ source files (if applicable)
• I have followed the style guidelines for Python source files (if applicable)
• I have made corresponding changes to the documentation (if applicable)
• I have added tests that prove my fix is effective or that my feature works (if applicable)