Don't use the same color for groups as trends

[snotskie]

Jenn suggested these changes in our demo meeting today:

1. In trend plots, add a legend, "-x", "+x", etc.
2. MeansRotation shouldn't use the exact same colors for groups as the trends use for direction, too much potential for confusion

I'm inclined to implement 1. first to see if that alone fixes the problem. 2. gets into a more general problem of "what colors look good for subtraction plots"

[snotskie]

Instead of using Cesar's colors, we could try ["#a4044d", "#454366", "#87c5af", "#d78b69", "#22695e", "#afb7dd", "#6a9012", "#be3acd", "#f6adff"]

For example:

Generated with http://vrl.cs.brown.edu/color and tested with https://projects.susielu.com/viz-palette

[snotskie]

[snotskie]

labels on x trend:

[snotskie]

compare with cesar's colors

[snotskie]

A radical idea is to always use the same colors for all trends and subtractions

That is at least legible regardless of what colors the groups get

[snotskie]

okay, I propose:

• Means rotation not inject the same colors as the trend lines
• All trends, subtractions included, use the same colors for the network edges
• We use cesar's categorical colors for groups

[snotskie]

implemented in https://github.com/snotskie/EpistemicNetworkAnalysis.jl/commit/4f3dd9583472e79150ecee99e3f71b5f7ca4799a