Fix some parameters after some optimization

[mb706]

Scenario: I optimize dimensions (a, b, c) for N steps, then determine that the value for c should be c*, but I want to go on optimizing over (a, b). Alternative scenario: I have done some optimizations over (a, b), on some instances of similar problems, indexed by a meta-feature "c". so I have a large table with columns (a, b, c, ) already. I start on a new problem where the the value of my meta-feature is c*, I want to tune using the prior info I have collected.

These two scenarios are very similar and should probably have a common solution.

[jakob-r]

Technical meaning:

• allow a different search space for the optim of the acq function with certain values fixed for the objective