Better handle scheduling large buckets

[domdomegg]

When scheduling e.g. 100 people at once, the algorithm can run out of time to find a good solution.

However, it succeeds when batching these up into e.g. 3x 33-person groups with roughly overlapping time availabilities.

We might be able to figure out how to improve the optimisation algorithm, or can implement a bit of a hack where if it fails to find a solution initially it splits up the group into smaller pieces.