Make Neff cut kwarg in `pipeline.hierarchical_likelihood`

[jaxengodfrey]

Currently, the minimum effective sample size is hard coded as Nobs. We should make this an optional argument that defaults to Nobs.