Use logical rather than physical navigation state to reconstruct G4NavigationHistory

[sethrj]

Related to https://github.com/celeritas-project/celeritas/issues/891: we need a higher accuracy way of reconstructing the Geant4 volume hierarchy from the Celeritas navigation state.

Here is how AdePT reconstructs the navigation index.

I added a touch of code in that direction: https://github.com/sethrj/celeritas/tree/navindex