Xmca

- bug fix of wrong results when predicting coslat weighted PCs

- resolve bugfix in predict
- resolve bugfix in hom/het patterns which were introduced after the upgrade from 0.x to 1.x
- add test cases, respectively, to avoid similar issues in the future

- resolve missing API in docs

- consistent storing location of plots
- add CI workflow

- method `predict` allows to project new, unseen data to obtain the corresponding PCs (works for standard, rotation and complex)
- more efficient storing/loading of files; Unfortunately, this and the point above made it necessary to change the code considerably. As a consequence, **loading models which were performed and saved using an older package version (0.x.y) is not supported.**
- add method to summarize performed analysis (`summary`)
- add method to return input fields
- improve docs (resolves 7)
- correct and consistent use of definition of loadings
- some bugfixes (e.g. resolves 12 )

- add arXiv pre-print for complex rotated MCA