generalizing to generic maximization problems

[tpapp]

This package defines an API for calculations involving log densities on ℝⁿ (mostly MCMC), with optional gradients (and Hessians), and an ability to query the dimension. The values returned by the log density are either finite, or -Inf.

But this describes unconstrained maximization problems as well, with -Inf being an infeasible set of coordinates.

I am opening this issue because I would like to use this API for the latter purpose too, but I am unsure about how to proceed. In order of increasing complexity and hassle, I can thing of the following:

1. we could document that it is OK to use this for maximization problems that do not actually involve densities, and not rename anything in the API,
2. rename everything accordingly using a more general name (to be bikeshedded)
3. provide an alternative API in parallel for maximization problems, essentially doing the same thing
4. in addition to 2. or 3., also rename the package
5. start a new package and retire this one.

[tpapp]

I would be inclined to do just 1.

The thought came up while refactoring TrustRegionMethods.jl, it would be so simple to use this API.