integrate maximum likelihood point capabilities

[lukashergt]

Some samplers (e.g. PolyChord) have settings that allow you to find the maximum likelihood point at the end of a run. It would be nice to incorporate that piece of information somehow into anesthetic.

This initially came up in #313 (comment).

Things worth considering:

• Is there a way of turning this maximum likelihood point into a final nested samples point (with a non-zero weight)?
• Alternatively, would it be simples to add it as a zero weight point?
• Could it be added with logL=max(logL) and logL_birth=max(logL)? Would that correctly determine the weight to be zero or nan?
• In how far would this need to be handled differently between MCMCSamples and NestedSamples?