Separating from #311, which hopefully is resolved in a more limited sense through #546. See also PR with initial work on implementation: #313. We may also consider allow this to be used in other calculations e.g. MD.
