Optimize Various Functions and Refactor

[Quantalabs]

1. Fix linting errors and format with black.
2. Uses a bidirectional BFS instead of a standard BFS, meaning the algorithm visits less nodes and completes faster.
3. Optimize margins matrix calculation in core by using vectorized operations as opposed to the previous nested for loop.
4. Refactor the impartial culture and Euclidean model, with slight efficiency updates (less nested for loops and more vectorized operations).

I'll try and get some benchmarks to see how much of a performance improvement there is.

[Quantalabs]

I added in the tests with 986a806, and the following is the output (running pytest --benchmark-min-rounds=100). I have:

• pytest
• pytest-benchmarks
• Python 3.12.4

BFS and DFS now have similar performance. I also reverted by core module to use the previous single-direction BFS implementation, which is the second image. The old BFS implementation performs worse on all measures.
New:

Old:

[Quantalabs]

Comparing the previous margins_from_ballots function to the new one, there is a significant increase in performance.

[Quantalabs]

I made a couple fixes but they don't seem to have affected the speed.

[Quantalabs]

It seems as though on large amounts of ballots, as in ba56f14, the difference between DFS and BFS is much clearer, and the bi-directional BFS is outperforming DFS.

[Quantalabs]

Finally, the tests for the improved models in b25eceb. All 3 perform better than their old counterparts.