Default weighting function normalization

[aywagner]

The default linear weighting function is being normalized per diagram rather than being fixed or normalized per experiment. As a consequence, the norm of a persistence image corresponding to a diagram with one point very close to the diagonal is not close to zero.

https://github.com/scikit-tda/persim/blob/76b2b5ae9d9d062fdf1117686c401e3855f3a05d/persim/images.py#L151-L160