Integrate the photoionization cross sections once for each balance

[barrettp]

Profiling a thick slab (one thread) showed that 57–64% of the time is in photoionization_integrals, and that the iteration of the electron fraction repeated it for every trial fraction.

• RatesMemo: for ParPhotoIonize1/2/3 the photoionization rate is independent of the electron density and the recombination rate is proportional to it, so the later trials of ionization_balance scale the first result. The rate matrices agree with the uncached ones to rounding (H, He, O, Fe).
• march_zones no longer computes line_emissivities twice when luminous=true (something(a, b) evaluated both).

one thread	before	after
thick slab, fixed T	73.8 s	51.6 s
thin slab, T iterated	166.6 s	133.3 s
balance of one zone	5.0 s	2.4 s

Results are unchanged (same He II to all digits; equilibrium T to 3e-10). Full suite passes on Julia 1.13 (987 tests). Further cost: record opacities, edge emissivities, heating and continuum opacity each still do a full photoionization pass per zone.

🤖 Generated with Claude Code