Density estimation interface

[kahaaga]

Describe the feature you'd like to have

As discussed in #277, we should have a long term goal of implementing a density estimation API, just as we have a probabilities API.

The reason is that the various (at least kNN-based) differential information estimators estimates densities. We should extract these density estimation steps explicitly, and place them in a unified framework, so that e.g. the differential entropy estimators become wrappers to an underlying density(est, x).

I'll have to check whether this can also be used on the order statistics based estimators.

[Datseris]

I don't think this should be in the 3.0 milestone, as the 3.0 should to be released sooner rather than later. This also does not sound like a breaking change, which is an argument against 3.0 milestone.