do not set ivar=0 to UNASSIGNED fibers by default

[julienguy]

Do not set ivar=0 to UNASSIGNED fibers by default because they might be useful for some analysis.
This addresses part of desisurveyops issue desihub/desisurveyops#76 .
(The other part being the counterintuitive default values of redrock redshifts for spectra with zero valid data)

[akremin]

A simple change that will partially solve the referenced issue. This does impact the masking for all pipeline steps, as it was implemented in a lower-level function. I believe this includes coadditions, so if we still want to mask these at the coadd/redshift level then we will have to do that in a future PR. I verified that this shouldn't impact sky modeling, because those are only selected among SKY targets, and this affects UNASSIGNED targets.

[sbailey]

For the record: redrock uses the fiberstatus UNASSIGNED, BADPOSITION, and BROKENFIBER bits to set the zwarn NODATA bit, so this PR allows the UNASSIGNED fibers to be fit but they will still receive a zwarn flag, which I think is what we want. This is similar to fitting redshifts for SKY fibers but setting the zwarn SKY bit.

Arguably it would be better to set the zwarn BAD_TARGET bit for unassigned fibers, but at least something is being set.