Reconsider approach to getting the full list of metrics, including relative metrics.

currently this is done in [some one-off R code](https://github.com/hubverse-org/hubPredEvalsData-docker/blob/145c4c82330acb32cf7cbb98d9e14395530137bc/scripts/create-predevals-data.R#L71-L95).  should we just assemble this list in javascript, or derive it from the column names of the data that is loaded?  See also https://github.com/hubverse-org/hubPredEvalsData/issues/4