add rise tabular

[cwmeijer]

Fixes #644, #712.

What works/is included:

• finds relevant features in simple test case
• should work fine with categorical (as well as numerical) features
• includes a single notebook (which fails because of the issue of the inconsistent outputs, described below; it is a 1-1 copy from the lime notebook which runs fine)
• adds the same tests for all 3 tabular xai methods

Things that are missing/broken:

• there are inconsistencies in the output types of the 3 xai for tabular data methods.
  • I think rise is working correctly as it uses the default return-only-class1 policy, see below (even though this is a strange default). See test_tabular on return shape.
  • Kernelshap seems to return a list of np arrays where it should a return a single array (with an added dimension for classes)
  • Lime returns all classes as an np array. This is what I was expecting with default parameters. However, I did see a mention of returning only class1 by default. see https://github.com/dianna-ai/dianna/blob/36c3aa21f678d73c8b5268dcabc1fe99a0a1f007/dianna/init.py#L134 I actually think this should be changed. No-one would expect this would be the default.
• links to the rise tabular notebook (in the readme etc) are not there