Changes in 2.1.0:
 - Function collapse() to generate
 pairwise graphs from vistla trees.
 - By default, i.e., when targets
 argument is missing, vistla assumes
 that it is a list of all features.
 Hence, the algorithm does not trace
 suboptimal paths (like ...->X->Y
 when ...->Z->Y is wider) anymore.
 These paths were rejected by most
 downstream processing, thus most
 users will only see faster processing.
 - The leaf_scores() function only
 made sense when suboptimal paths
 were traced, so it got removed.
 The usefulness of it was limited,
 though, it is a better idea to use
 ensembles to analyse alternative paths.
 - Package vignette.

Changes in 2.0.4:
 - Documentation updates following
 publication of the method's paper.

Changes in 2.0.3:
 - Fixed the flow generator, which used
 to mix `from` with `into` constraint.
 - Updates to the example data sets.
 - Portable version of the algorithm
 exposed as a C library.

Changes in 2.0.2:
 - Changes in the example datasets to 
 make documentation more consistent.
 - Ensemble option to get a consensus
 over many iterations of vistla.
 - Better handling of hierarchy objects,
 which are also a result of ensemble;
 in particular, path_to(), paths(), and
 prune() work on those objects.
 - Changes to the internal PRNG, old 
 results for certain seeds may change.

Changes in 2.0.1:
 - Fixed CRAN warning (irrelevant to the
 package functionality).
 - Error is emitted when the number of
 objects is too large for KT to function.

Changes in 2.0.0:
 - Information estimation argument; vistla
 no longer performs auto-discretisation of
 continuous inputs.
 - Random tie breaking works properly.
 - Plots are more compatible with the wider
 grid ecosystem.
 - Integrated Kendall transformation for
 processing of continuous variables.
 - agreement() removed.
 - Test suite only runs on CI.

Changes in 1.1.0:
 - Random tie breaking.
 - Grid-based plot with improved layout.
 - More verbose flow control.

Changes in 1.0.1:
 - PROTECT order bug fixed.

Changes in 1.0.0:
 - First public release.