Order statistics and ordered samples

For linearly-ordered domains (especially ℝ), we often need to compute measures like
- the `r`th [order statistic](https://en.wikipedia.org/wiki/Order_statistic) from a sample of size `n`
- an ordered sample of size `n`

These are very different cases. The first case is a single sample, the `logdensity` is given by

For example, it would be convenient to be able to write

```julia
OrderStatistic(100, 5, Normal())
```

For the second, the sample is of size `n`. Here we probably want something like what Stan does,
https://mc-stan.org/docs/2_26/reference-manual/ordered-vector.html