Improve neighbour-joining tree performance

[alimanfoo]

• Use anjl instead of biotite to build neighbour-joining trees, which is capable of building trees for 20,000+ samples in around 5 minutes. Resolves Neighbour-joining tree performance #627.
• Cache the results of tree construction to avoid repeated computation.
• Use anjl for tree plotting.
• Also work towards Anopheles refactor #366.
• Along the way, fix issues encountered when plotting and data includes missing values.

---

• Create a new module malariagen_data.anoph.distance.
• Add a class AnophelesDistanceAnalysis.
• Move the public method biallelic_pairwise_distances() and associated private methods from the AnophelesDataResource class to the new AnophelesDistanceAnalysis class.
• Add a public method njt() to the new class which returns an array Z and which also saves results to the results cache.
• Add a private method _njt() which is called from within njt() to run the actual neighbour-joining if the result is not cached, calling biallelic_pairwise_distances() internally to obtain a distance matrix then anjl.rapid_nj() to compute a tree.
• Add a public method plot_njt() to the new class which calls njt() first to obtain a tree then calls anjl.plot() to create a plot.
• Remove the plot_njt() method from the AnophelesDataResource class.
• Add AnophelesDistanceAnalysis as a parent class of AnophelesDataResource.
• Add tests to get coverage of the new AnophelesDistanceAnalysis class.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[leehart]

Awesome. Thanks @alimanfoo 👍