Allow grouping of heatmap elements by metadata

[polyatail]

This PR addresses the following issues:

#136 I broke the box plot and dendrogram plotting functions out into their own viz/_primitives.py file. This will make it easier to re-use these bits of code.
#151 By user request, plot_heatmap now takes a categorical metadata field (or tuple containing multiple metadata fields), and performs clustering (with user-specified metric and linkage) on a per-category basis.

This code will generate clustered heatmaps at each taxonomic rank:

ranks = ['species', 'genus', 'family', 'class', 'order', 'phylum']
plts = {}

for r in ranks:
    plts[r[0]] = samples.plot_heatmap(
        metric='weighted_unifrac', linkage='median',
        haxis='fish', threshold=0.1, rank=r,
        return_chart=True
    )

plts['p'] & plts['o'] & plts['c'] & plts['f'] & plts['g'] & plts['s']

In the below example plot, the value of metadata field 'fish' is reported above the heatmap.

[coveralls]

Coverage increased (+0.05%) to 84.038% when pulling 35fbe9f on roo-kiwi3 into 78f9dca on master.

[boydgreenfield]

Why does this decrease test coverage in _heatmap.py by 50%? Can we add a test to address that?

[polyatail]

I will, I have an issue open for tests and was gonna do it there

…

On Mon, Jan 14, 2019 at 10:14 PM Nick Greenfield ***@***.***> wrote: Why does this decrease test coverage in _heatmap.py by 50%? Can we add a test to address that? [image: image] <https://user-images.githubusercontent.com/535969/51162072-bd5a4600-1849-11e9-846d-f31474d29649.png> — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#172 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAv3BKzQAsswLnSILO-k54mRYsmqad59ks5vDXHCgaJpZM4Z_8rz> .

[boydgreenfield]

@polyatail Works for me!