Optimise the boids algorithm

Options:
- store sorted fish between frames (linear improvement to a O(n^2) time complexity)
- store fish in a 3d locality datastructure (octree or k-d tree) (sub O(n^2))
- cache the 3d locality between frames (linear improvement to sub O(n^2))
- submit the job to a GPU compute shader

Start by locating the boids simulation implementation and measuring its current performance. Compare the listed approaches—sorted fish, a 3D locality data structure, cached locality, or a GPU compute shader—and define which optimization to pursue. Done should include a demonstrated performance improvement without changing the simulation’s intended behavior.

Written by the indexing model from the issue text.