Improve docstrings in `InversePropensityWeighting` class

[drbenvincent]

There is scope to improve the docstrings in the InversePropensityWeighting class. While the class does have good docstring coverage, and describes the purpose of the method, it does not always describe what the args/kwargs are.

Tagging @NathanielF, if you're interested?

NOTE: Do this after #381 is merged.