Fix n-1 parameters for distributions such as lnorm with MLEs

Expected to provide a speed gain (could examine)
Loses SE on parameter estimate which not a problem as get CIs by bootstrapping
Requires at least one parameter estimate to be optimized.
Not a general solution for censored data (appears to be ok for invpareto which uses max of right but would fail if right = Inf)

Start by locating the distribution-fitting entry points for lnorm and other distributions that use MLEs, then inspect how n-1 parameters and censored data are handled. The issue notes expected speed gains, lost standard errors, an optimization requirement, and an invpareto edge case; done should address the intended cases without claiming a general censored-data solution.

Written by the indexing model from the issue text.