Normalization off and some PDFs are zero on entire grid

[sschmidt23]

It appears that the output pdf normalization if off by a factor of the grid spacing. Also, for "bad" PDFs where none of the templates are a good fit and chi^2 is very large, the likelihood is zero everywhere. Having all zeros can be bad for normalization, we likely should replace this with a normalized uniform distribution (which is accurate, we have no idea what redshift is best, so all are equally likely)