Double-check LPL covariate structure

[swo]

Ensure https://github.com/CDCgov/cfa-immunization-uptake-projection/blob/5ef92e2e6d8cb0443e8f61265faca6298578a75a/iup/models.py#L167-L168 does what we think it should

[swo]

The use of np. here also means that we don't get JIT, which I presume is slowly down the fits substantially

[swo]

I experimented with this in #229; see notes there.