Running the plot_njt() function on larger datasets can take upwards of 20 minutes or more specifically in the stage after the pairwise distances have been computed and during the construction of the neighbour-joining tree..

Perhaps there is an opportunity to accelerate or parallelise the neighbour-joining tree construction. Currently we use the implementation from biotite, which was faster than either BioPython or scikit-bio, but perhaps there is room for further improvement.

We could also at least cache the results of the tree computation so we don't need to rerun it.