Add cluster width as a function of eta plots to IDPVM
Actually, this also changes the InDetPerfPlot_HitResidual class to not use raw C arrays for storing histogram pointers, instead using std::vector. I think this is more robust (I have a suspicion that the previous implementation was relying on UB), and I don't think the bounds checking should be relevant in terms of performance.
I'm also adding new (old) plots: cluster eta and phi width as a function of eta for pixel, and phi width only for SCT.
The ITk specific XML file had some references to a previous implementation of this that I couldn't find (maybe 21.9?), which I removed in favor of the central common XML.