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 commit acf483a1

Split the model comparison to be callable in functions for bayes factors, model weights and justifiability analysis and renamed the overall function

BMC only runs the BME part of BayesInf, split JA into its own function, removed redundant function in MCMC, added to function diagrams

More test for BayesInf, starting tests for MCMC

Some simple fixes for various test, start on testing BayesInf, output directory as changeable parameter in BayesInf

UMBridge examples cleaned and pylink 'single-valued' removed

Solved some issues from rebasing

OLS constraint directly to be set in the MetaModel source code. Small regression comparison as example for the constraint. No unit tests yet for the feature

Added a minimal test based on pylink classification. More work needed for higher-dim. model outputs here

In this scope the multiprocessing for running the models does not seem to work correctly anymore

Got started on these two tasks, restructuring still needs a lot of work here

Code updated for static training, tests not done yet

New class 'Engine' responsible for both training sequentially and non-sequentially.
Currently training with old MetaModEngine and in MetaMod directly still possible, but will be deprecated and removed soon.
Also started splitting ExpDesign into InputSpace (for MetaMod, created dists and polynomials) and ExpDesign (for Engine, can also sample)

Started testing MetaModel while also refactoring a bit more by moving code between ExpDesign and MetaModel

Added tests for Inputs and started on those for ExpDesign.
Regarding the second there are some open questions noted in issue #32

Fixed the bug in the sequential design.
Typo correction also added here.

Implemented simple example using the tsunami model from UMBridge