Performance enhancement

Replace cvmix_get_kpp_real calls with direct access of datatype:

1. Testing in POP showed no answer changes
2. KPP is ~10% faster than v0.92-beta, and POP overall (with iage as the only tracer module) was ~4% faster
3. POP still runs fastest using native KPP instead of CVMix
   • CVMix is ~60% hit on the KPP runtime and 15% on all of POP (down from 80% and 20%, respectively)