Support nonuniform energy grid present in Geant4 11.1

[sethrj]

In geant4@10.7:11.0, EM cross sections are stored on two separate physics tables that share the same log-energy spacing, but the second table stores $E_i \sigma_i$ rather than $\sigma_i$. Unfortunately geant4@11.1 adjusts the lower energy grid (I think based on the energy cutoffs), so that the grid spacing is no longer the same for the two tables. Our assertion does catch this in debug mode:

/home/s3j/.local/src/celeritas/app/demo-loop/demo-loop.cc:244: critical: While running input at <stdin>: /home/s3j/.local/src/celeritas/src/celeritas/grid/ValueGridBuilder.cc:111:
celeritas: precondition failed: has_same_log_spacing(lambda_energy, lambda_prim_energy)

We will have to either change the xs grid to support two different spacings (which should probably be rolled into the refactor from #253) or completely recalculate the cross sections ourselves (which is what G4HepEm does).

[sethrj]

This is due to Geant4 change set phys-ctor-em-V11-00-26: fixed gamma general at low energy (V Ivanchenko):

2022-07-06 V. Ivanchenko (emlowen-V11-00-08)

• G4Rayleigh scattering moved boundary between low and high energy
  table from 100 keV to 150 keV to be coherent with G4GammaGeneralProcess

And G4GammaGeneralProcess has a "low photoelectric" boundary of 150 keV.

Here are the cross sections compared from Geant4@11.0 and @11.1:

So we should be able to work around this at least by turning off Rayleigh scattering?

[sethrj]

Honestly I'm surprised this is using a uniform log energy grid at all; that bump that looks like a possible resonance is indeed a resonance: