v2.0.0

Total refactor with minor breaking changes

• Code is now way faster, changed the permutations from a re-indexing/shuffle scheme to a matrix-vector product scheme that is much more compute friendly. In some cases its orders of magnitude faster
• The linearization scheme has been totally revamped, now uses a rectnagular distance matrix (default is the full square matrix which has quadratic compute complexity). Can get several orders of magnitude speedup in heavy (large sample size) runs.
• PTED now includes a containment test that checks is x is covered/contained in the y distribution. This is an asymmetric test that isn't checking for equality.
• A lot more tests check that everything works properly

Breaking changes

• The chunk_size and related arguments are no longer available. Now use the n_landmarks system
• The energy distance is computed without the diagonal elements, produces a slightly different metric. Does not affect p-values
• The energy distance statistic is scaled by n1 * n2 / n . Does not affect p-values

What's Changed

• Convert to landmark rectangular matrix system, also big speedup by @ConnorStoneAstro in #21
• improve workflow stability by @ConnorStoneAstro in #22

Full Changelog: v1.7.0...v2.0.0