Idea: using vega-lite for bedstat plots instead of producing them with ggplot

[nsheff]

Here's an idea:

What if you could get the individual statistics behind the GenomicDistribtuion plots via an API?
Maybe we don't even need to produce the PDF/PNG plots using ggplot, and instead, we just store the underlying data, and then use vega-lite to make plots, while also making those data points accessible via API.

So, the advantage would be:

• No need to produce and store plots
• A user could hit an endpoint to get very detailed statistics about a region set of interest

I guess this is kind of already what we're doing with the feature overlap % plots, where you can query the % exons, etc, as independent values. So, maybe what I'm proposing is to extend this.

Probably there's a happy medium somewhere -- it doesn't make sense to make APIs around some of the plots that make big distributions of lots of little points, as we don't want to store every point. But, summaries of those could be useful. Summaries wouldn't be able to re-create the entire plot. So, there's room for both approaches.