Add PZ widths systematics model

[slosar]

At the moment there are only PZ shifts but no PZ widths in firecrown/ccl/systematics/pz.py.
PZ widths seems like an easy thing to add. Long term we want a better way modelling PZ uncertainity, including potentially using samples of N(z) and/or eigen-vector like marginalization.

Let's make PZ responsible so that they feel so bad about it that they will implement something better.

[slosar]

This is a very specific request, c.f. more general ticket: #50

[rmandelb]

The long-term more general parametrization could be explored as part of https://confluence.slac.stanford.edu/display/LSSTDESC/Requirements+on+PZ+errors+for+3x2pt ?

[beckermr]

Yes! I was wondering if @empEvil was working on this...

[marcpaterno]

The current implementation has PhotoZShift in both weak_lensing.py and number_counts.py. These should be extracted and put into a single place where they can be used by both modules, perhaps into a more general dn/dz class to be used by both weak lensing and number counts.

Is anyone from the PZ working group interested in adding the width effects? Would the same proposed dn/dz have a way to introduce the width effects that would be useful to both weak lensing and number counts?

Adding @bruno-moraes to the discussion participants.

[tilmantroester]

Having a single PhotoZ systematic for both weak lensing and clustering would be useful I think. Adding width scaling on top of the existing shift systematic can be done quite easily.
There's also ongoing work on a more general framework to get n(z) uncertainties from RAIL into firecrown.