mcse for minimum / maximum

[wds15]

It could be a useful feature to have an mcse for the minimum & maximum of a posterior distribution. This is not trivial as it requires a definition of what the tail (beyond the extremum) looks like. Maybe @avehtari can add some details as to what would be needed. However, before starting this we should also find good use cases. This issue is a follow-up from the issue on the inconsistent quantile function behaviour.