add ego entry point (Issue #138)

[jakob-r]

No description provided.

[berndbischl]

lets us try to specify the API here:

1. remove init.des.points. if not, the user has to tell the system HOW to gen the init design. blows up arg signature. let the user pass the des, if he does not, do a reasonable init des. i would suggest maxminLHS with 4*d.
2. we need more simple term conditions. budget, iters, and a bound on the y value to reach.
3. you need to check that parset only contains numerical stuff
4. i dont think you can let the user pass a control object. i could use that do request "weird" stuff, non-ego stuff. if you want that, the full control, use mbo.
5. its not "visible" how the learner is constructed. at least link to some doc page for this.

@mllg ?

[jakob-r]

For 5 see #223

[mllg]

We now have makeMBOLearner which approaches this very problem from a different angle. IMHO the construction of a default learner takes away most of the pain. I guess you cannot remove the burden of some choices from the user, like the termination criteria. Unsure how we should proceed.

I guess I would rather favor to postpone this one for now.

[berndbischl]

I agree that this should be postponed!

But: makeMboLearner simply constructs a reasonable default model. This PR is about something else. And more general. Lets not mix the two.