[ ] ~To get the code coverage to 100%, there must be a test of conditional MI with k=N.~
Driver requires k<N, I don't want to introduce special code for this test only.
[x] There is not yet a driver test of conditional MI with multidimensional condition. (#42)
[x] More distributions with analytically known MI, see Darbellay & Vajda (DOI: 10.1109/18.825848) and Kraskov et al. (DOI: 10.1103/PhysRevE.69.066138) (#42)
[x] The "too large k after mask" test does not take lags into account. (#38)
[x] Other corner cases with confusing error messages? (#38)
k=N.~k<N, I don't want to introduce special code for this test only.kafter mask" test does not take lags into account. (#38)