Support multiple elements per material

[sethrj]

Currently we're using a simplification of allowing only one element per material, so that we don't have to calculate/tabulate/sample elemental cross sections in a given material.

Our CERN collaborators informed us that many of the Geant4 physics models actually use tabulated-by-element cross section data rather than calculating on the fly, so we should be able to build element sampling (in most cases) into the process/model selection kernel.

• [x] Export tabulated micro cross sections from Geant4 (#326 )
• [x] Preprocess micros into CDFs/alias tables during Celeritas setup
• [x] Sample elements after model selection in the "discrete" kernel (for those that have tabulated cross sections)
• [x] On-the-fly element selection for OTF processes (photoelectric, annihilation) inside the interaction kernels

[sethrj]

See also https://github.com/celeritas-project/celeritas/issues/60 where we discussed the first aspects of element/process cross section calculation. I probably should have read it over more carefully before closing it and opening this guy 🙄

[sethrj]

This was closed by mistake 😬 still work to be done.

[sethrj]

Closed by #440, thanks @amandalund !