Training on constant values

[mb706]

Different calls with constant response seem to not work well. E.g.

DiceKriging::km(design = data.frame(a = 1:3), response = rep(1, 3))
DiceKriging::km(design = data.frame(a = 1:10), response = rep(1, 10), optim.method = "gen", nugget = 0)

both give errors.

This came up in mlr-org/mlrMBO#490