Additional acquisition functions for sequence model-based optimization

[SimonBlanke]

I looked into a way to add more acquisition functions for the sequence model-based optimization algorithms. In the current version 1.0 the only acquisition function is the expected improvement.

Since the xi-parameter is specific to the expected improvement and currently part of the Bayesian Optimizer API it will be necessary to introduce breaking changes to the API. The xi-parameter must be removed in general and should only be passed if the expected improvement is selected. Other acquisition functions might not have the xi-paramter. All of this means, that this new feature should be part of v2.0, because there will be a small API change (because of the xi-paramter).

I also started to clean up the code for the expected improvement and put it into a separate module in: cd6747c33e14e638faa8d3368e70f2c3064fdfb2 The future acquisition functions will be part of this module.