Simplified and extensible interface, additional BO techniques, inverse modelling

This release aims to facilitate getting started with SOBER by providing a central interface that easily allows using all the customization options with SOBER without unexpected bugs. The same interface allows for extension of SOBER with popular Bayesian Optimization techniques with minimal code. The following techniques have been added using this interface:

• Expectation Propagation
• BOLFI
• inverse modelling

v1.1

Additions:

• Set up packaging infrastructure to produce install files.

Changes:

• Made device and dtype settings changeable during program execution.

Fixes:

• Fixed some minor issues that prevented non-PyTorch target functions from working.