Add SVARM approx Notebook

[mmschlk]

Add an example notebook describing how the SVARM approximator works and what it does.

Also explain how it relates to SVARM-IQ.

In the end a small benchmark comparison with KernelSHAP and Permutation Sampling would be nice.