ENH: Add `augur colors` subcommand based on `assign-colors.py` scripts

[corneliusroemer]

We discussed adding more reusable components to augur at our meeting yesterday.

Working on ncov PR https://github.com/nextstrain/ncov/pull/1050 I encountered some limitations of the assign-colors.py script. This script could benefit from new features.

The script is used in a lot of places, it could be a prime candidate for augurification: see where it is used https://github.com/search?type=code&auto_enroll=true&q=assign-colors.py

The ncov version is here: https://github.com/nextstrain/ncov/blob/master/scripts/assign-colors.py

Some features to possibly add to the existing script:

• Ability to ignore metadata (columns) for certain colorings
• Ability to use node data jsons in place of metadata (particularly useful for clades jsons)

[corneliusroemer]

Just rediscovered that augur export v2 doesn't accept color_ordering.tsv despite this being a pervasive config file pattern for current workflows. Another argument for augur colors

[huddlej]

I just reinvented the wheel a bit in the seasonal flu repo with a similar command to generate a custom color scale in an Auspice config JSON. Thanks to @joverlee521 for reminding me that this is a more general need across pathogen workflows!

I can see how it's useful to get different output types like an Auspice config JSON color scale or a colors TSV file, depending on what you're trying to do. It's also useful to limit the values that get assigned colors to a subset of what appears in a given analysis. One could imagine supporting both an opt-in approach like the script I linked to above where the user specifies the values that should get colors and an opt-out approach where the user specifies the values that should not get assigned colors.

[joverlee521]

During discussion today, we floated the idea of vendoring the color script similar to how we vendor nextstrain/ingest scripts since assign-colors.py feels a little too ad-hoc to officially release as an augur command.