Implement different types of encoding for Voxel::Detail

[Kurble]

Currently the Voxel::Detail variant only supports a plain array for storing sub-voxels, but this is not scalable if your voxels contain a lot of per-voxel data. The current array implementation is also not really cache coherent if you access a lot of neighbouring voxels.
A solution is to implement a system that allows for different types of encoding. An Encoding trait should allow iteration over the subvoxels of an encoded Voxel::Detail node. It would also be nice to stack different kinds of encoding, so that you could apply for example both run length encoding and a morton order.

Todo:

• Define the Encoding trait.

We want at least three types of encoding:

• Plain array encoding (part of this issue)
• Run length encoding (issue Implement run length encoding #10)
• Morton encoding (issue Implement morton ordering. #11)