Skip atomic relaxation emissions below the photon energy cutoff

[GuySten]

Description

After a photoelectric or Compton interaction, atomic relaxation follows the full vacancy cascade down to the outermost shells, sampling a transition and an isotropic direction (with sin, cos and sqrt) at every step. For heavy elements most of this work has no effect on the simulation:

• fluorescence photons below the photon energy cutoff are never created (create_secondary returns before banking them), and
• Auger electrons below the photon energy cutoff cannot produce bremsstrahlung photons above it (thick_target_bremsstrahlung returns immediately), and with local energy deposition electrons produce nothing at all.

Every particle emitted while filling a vacancy, and every vacancy left behind, has an energy below the binding energy of the vacancy shell. This PR therefore:

1. skips vacancies whose binding energy is below the photon energy cutoff, together with the rest of their cascade, and
2. for the remaining vacancies, samples a direction and creates or processes the emitted particle only when its energy is at or above the cutoff.

The physics is unchanged, but the random number sequence changes for any problem with atomic relaxation, so some photon regression results need to be updated.

For W and Pb with the default 1 keV photon cutoff, roughly 80–85% of the vacancies in a K- or L-shell cascade lie below the cutoff (for example, a Pb K-shell cascade visits about 23 vacancies, of which about 3.5 are above 1 keV). Profiling the gamma shielding model from this Discourse thread (Cs-137 source in a W/Pb cask, analog, single thread) showed atomic relaxation taking 26% of the run time. With this change, transport time dropped by about 22% (22.4 s to 17.4 s for 10⁶ histories), and the tally results agreed with the original within statistical uncertainty.

I also added a paragraph describing this to the atomic relaxation section of the photon physics methods documentation.

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)