Consider refactoring "old" np.dot and dot() operations by "new" @ operator

[BjoernLudwigPTB]

Since Python 3.5 there is a new and for higher dimensions semantically more intuitive version of a matrix multiplication operator @. We usually deal with 2-D inputs, when we use the dot-functions, but we might easily move to higher dimensional operations by simply using the @ and would be more on the brink of Python developments.

We should thus consider shifting to @.