Refactored functions to evaluate/check metric and eval time arguments

[topepo]

This includes reworked versions of the argument checking code used by a family of functions: fit_resamples(), tune_*(), and last_fit().

I've wholesale copied parsnip::.filter_eval_time to use here; we should probably export that.

For testing tune_bayes(), we'll need an additional engine so that will have to go into extratests.

An updated itinerary of PRs

1. Tools for selecting a default evaluation time #768
2. Refactor choose_metric #777
3. Refactored functions to select single evaluation times #778
4. (you are here) refactor metric input checking code in the tune_* functions, fit_resamples(). etc.
5. update finetune to use these changes
6. put unit tests that can't go into tune in the extratests repo
7. add additional functions that return (potentially) multiple evaluation times (eg autoplot())
8. write an article on tidymodels.org about this.

[hfrick]

Couple of questions about future directions, collected from the comments:

• what is the plan for for check_metrics() vs check_metrics_arg() and check_eval_time() vs check_eval_time_arg()?
• are we still aiming to remove tune's dependency on censored?
• we could include parsnip's .filter_eval_time() in the standalone file for survival utilities
• .get_metric_time() is now unused - let's remove/deprecate it?

A consistency comment

• warn instead of error on <2 time points for integrated survival metric?

An oversight

• the snapshot test for last_fit needs to be properly updated

[github-actions]

This pull request has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.