[Meta] Implement an ideal library graphs, shared components, and statistical methods

[brylie]

This task is to plan and track progress towards a baseline library of useful graphs. It also provides an opportunity to unify and plan chart components, so they are shared throughout the library.

Chart library

Metrics Graphics currently supports the following chart types:

• [x] Line
• [x] Histogram
• [x] Scatter plot
• [x] Bar
• [x] Data table

Additionally, the following chart types may be generally useful¹:

• [ ] Bullet Graph
• [ ] Dot plot
• [ ] Sparkline
• [ ] Box plot (#518)
• [ ] Spatial map
• [ ] Heat map
• [ ] Tree map

  Shared components

• [x] Axis label
• [x] Axis scale
• [ ] Trend line
  • [x] Least mean square
  • [ ] Local regression (LOWESS/LOESS) #540
• [ ] Color palette

  Reference

  1. Few, S. (2013). Information dashboard design: Displaying data for at-a-glance monitoring (2nd ed., pp. 119-150). Burlingame, Calif.: Analytics Press.

[hamilton]

Great issue, and good use of the checkmark stuff. I think for 2.8 our main goal is going to be code cleanup, with a focus on scatterplots once that task is done. The scatterplot implementation is a bit boilerplate-y, and we are not currently reusing a number of useful components from line.js. Same with histograms and bars - woefully undesigned as of now. Verbally @almossawi and I have essentially only considered line charts to be somewhat feature-complete and stable.

With this in mind, I think the best way to begin this process would be to tackle the whole of #534, component by component, and get everything up to snuff, code-wise. From there we can reuse a number of components from the line chart implementation for scatters (specifically elements like rollovers) and clean up many of the scatter-relevant features, such as least squares (which is not thoroughly tested) and a loess implementation.