NEO8: account for larger typical number of unWISE coadd epochs

[ameisner]

Post-DR10, we would like to photometer the new NEO8 time-resolved unWISE coadds, which include ~9 years of data and therefore typically have 18 coadd epochs per band per unWISE tile footprint. For DR10 (NEO7), the typical number of coadd epochs per band per tile footprint was 16. I believe that accommodating this increase in the number of WISE light curve epochs will involve small changes in legacypipe, to increase the size of the arrays in the relevant WISE light curve Tractor columns and also to have the ecliptic pole epoch subselection aim for a subset of 19 epochs rather than 17.

This issue is analogous to #693 from summer of 2021.

[ameisner]

I'm assuming that I should wait to do this until DR10 processing is over with?

[moustakas]

@ameisner, yes, let's please wait before making this change. Although DR10 is using a fixed container, I am occasionally using legacypipe/main this summer but am not planning to switch over to NEO8 just yet. Thanks!