Use unordened sets where better

Currently we only use (ordered) sets.

Unordened sets in C++ have a better performance (amortised constant cost) for retrieving a single element (the .find(X) operation). For iterating over subsets (the .cbegin iterator etc.) however, it has a higher cost.

It might be worthwile to see if there are sets where we only use the .find operation and use unordened sets for these to improve performance.