Removing console output from ehvi + Slight code refactoring in experimental + Moved acquiopt tag from bo_multi.hpp since it seems to be algorithm specific

- sferes variant fails on execution but compiles
- create a new separate stat header for the experiment (same code as before)
- Code refactoring

explicitly call the HP optimization + add a parameter in bayesian_optimizer to control the frequency of the HP optimization

Bug: when calling the default constructor of gp and using multidimensional inputs/outputs, raises an assert from eigen. Corrected by allowing the gp to autonomously determine the dim