Fix judicious partitioning runtime/memory consumption

[adrianzap]

Find a way to avoid the mad memory consumption of the T set so we can use it on non-tiny datasets.

Possible strategies:
-find another algorithm for the set cover problem
-just calculate a part of T --> check the paper for this (might not be possible)
-don't calculate T at all but use some implicit representation of it (probably not feasible)
-use hypergraph coarsening (probably memory consumption will still be far too high)