Performance improvements

[clarenc3]

Some of these will be addressed in NUISANCE 3, but leaving a list here in case

• Make splines on-the-fly
• Check particle loops; GetHMFSblabla is nice for the user, but we end up looping over the same particle stack many many times
• Is the signal definition checked twice...?
• Load events into RAM rather than TTree::GetEntry; should be relatively large improvement with not much human cost

[KSkwarczynski]

After reading this issue I got really suprised Nuisancedoesn't use stuff like multithreading or simd.

I assume this is moslty due to TTree::GetEntry and reweighting eninges not being thread safe