Infrastructure for superpositions

Users can build a `SuperpositionMeasure` in a few ways. We need `a + b` and `sum((a,b))` to result in the same thing. I think we should have one method to handle both cases. Whether that's a method on Tuples or a `...` method mostly comes down to performance and which is more natural to code.

There's a fair amount more to do for general superpositions. In the end we need a data structure that encodes how the base measures are related, storing everything in a tree structure. That will make it easy to compute things efficiently, since it will let us share log-density computations. It will also make it easier to fold more measures into an existing superposition.

We'll also need to build the `logmass` API we've talked about before.

_Originally posted by @cscherrer in https://github.com/cscherrer/MeasureBase.jl/pull/46#discussion_r863853741_