check metrics when adding results

[topepo]

We now check after the fact when the metrics are used (e.g. autoplot()). We might be able to figure this out before the model fits and then add a warning.

We should also use the upcoming rsample fingerprinting method to make sure that the same resamples are used.