run unit test for tuning with deps in defaults

[berndbischl]

i think the issue reported here is important enough to warrant a unit test with the usecase for tuning

https://github.com/mlr-org/paradox/issues/259

[jakob-r]

For completeness I will repeat what I said here

The defaults are just a documentation (this was kind of our understanding IMHO):

• We cannot ensure that they are correct
• We never pass them explicitly to the learner
• ==> They are not reliable and should not be coded against.

If you want to tune a param that depends on the default just follow the very helpful error message and set the parameter manually in lrn$param_set$values.

[pat-s]

If you want to tune a param that depends on the default just follow the very helpful error message and set the parameter manually in lrn$param_set$values.

Haven't read all of your discussion in #259 but is that the final statement you agreed on? If so, I would pass this as a solution to the initial question. Or is this still in discussion?

[be-marc]

We do not check this anymore by default. This can be activated with check_values = TRUE. In this case the dependency must be set. Unit tests in #440.