All combinations of tradeoff-exploration-exploitation run through with SeqDesign. Reverted BayesInf to the current develop branch

example_analytical_function_testSequential runs through all the sequential options and contains notes on which do not work.
Choose_next sample now uses only one set of candidate samples

User guide startet for models, surrogate training, sequential training, postprocessing and BAyesian classes. The guide now contains simplified UMLs (for BayesInf not yet done).

Webpage includes quickstart, user guide, tutorial, example collection and api.
One new example was added for the user guide.

An example 'principal_component_analysis' was added to specifically test this functionality. It uses the analytical model.
Fixed two if-calls in `pca_transformation` to match older code and the content of the function.

This merge is to be a part of release 1.1.0.
It contains bug fixes and restructuring of the classes BayesInference and BayesianModelComparison, as well as new tests and examples and the first draft of a package-specific website.

Example 'only-model' created based on the 'analytical-function' to test Inference without a surrogate.
Currently it's not running without building the surrogate - do tests first and see if able to refracture?

User guide startet for models, surrogate training, sequential training, postprocessing and BAyesian classes. The guide now contains simplified UMLs (for BayesInf not yet done).

Webpage includes quickstart, user guide, tutorial, example collection and api.
One new example was added for the user guide.

An example 'principal_component_analysis' was added to specifically test this functionality. It uses the analytical model.
Fixed two if-calls in `pca_transformation` to match older code and the content of the function.

MCMC does not receive the whole BayesInf object anymore, performance of inference now should be back up to speed with MCMC. In future perhaps rework so that MCMC is a child class of BayesInf?

This reverts merge request !18