Replace weighted average with general kinetic model for multi-PLD data

[tsalo]

Summary

This was originally part of #268, but I wanted to get that merged ASAP, so I am happy to deal with this issue in a later PR. Basically, I need to decide between the Juttukonda 2021 unweighted average and the Dai 2012 weighted average for multi-PLD PCASL CBF calculation (or another approach entirely).

[tsalo]

Manuel Taso provided some very useful advice recently. Namely, we should only use the weighted-delay approach when vascular crushing is enabled. When the scan doesn't have vascular crushing, this approach could cause CBF over-estimation.

Instead, Manuel recommends using a 2-3 compartment model (with an intravascular compartment), which can fit arterial bolus arrival time, transit time, and CBF all at once, as recommended in this preprint: https://osf.io/4tskr

[tsalo]

It looks like I really need to implement the general kinetic model.