M-LOOP now some new options and has a different default behaviour.

• A differential evolution algorithm has been added. This can now be used on its own or as a trainer for the gaussian process (GP). It is now the default trainer for the GP.
• Added a shell based interface.
• Improved the documentation. Added a tutorial for using M-LOOP as a python API and also improved the installation instructions. New features have also been appropriately documented.