Include execution time as a metric for workflow sets

[luisDVA]

Feature

This came up on Day 1 of the ML w/ tidymodels workshop at rstudioconf.

In some cases it may be useful to consider code execution time as a metric for ranking models in a set in addition to the other metrics. Davis mentioned that intermediate parsnip objects store execution time so this can be extracted and used to weight predictive power with running time. Thanks!

[simonpcouch]

This is an interesting idea! We'll come back to this after rstudio::conf. :)

[simonpcouch]

Related to tidymodels/tune#248! Some work starting to come together here.