Fix random delaunay triangulations to not violate Euler characteristic

[acgetchell]

We should add tests for the d-dimensional Euler characteristic.

Then fix the randomized delaunay triangulation generators to only produce correct triangulations based on Euler characteristic.

Eventually we'll need topology.

[acgetchell]

This issue depends on canonical topology traversal (edges / adjacency) and will be re-evaluated after the public topology API lands in v0.7.0.