Improve tests

[graeme-a-stewart]

There is a nice benchmark of 100 Pythia 8 events that @grasph generated that we use for correctness and benchmarking ATM.

I would propose that we extend this to two additional use cases:

• low density FCCee type events (where N2Plain might outperform N2Tiled)
• very high density HI type events (which will just be challenging)

In both cases FastJet should be used as a reference to ensure correctness.

We should also test other clustering strategies, at least Cambridge/Achen ($p=0$) and Inclusive $k_T$ ($p=1$).

[graeme-a-stewart]

This was fixed in 47182d9. All 3 main algorithms are tested, as well as different input types: PseudoJet and LorentzVector.