Generate a chart of the "best case" network coverage, based on nodes' claimed radii

[carver]

Stacked area chart, showing network replication (based on the claimed radius of the node)

### Tasks
- [x] Paper protototype of how xor distance plays out in a replication chart
- [x] Generate a stacked area chart in d3 (or maybe just a bar chart), using dummy data
- [x] Calculate chart data, based on node ID & radius data
- [x] Test tradeoffs between high-resolution/low-speed and low-resolution/high-speed display
- [x] Clean up for merge
- [x] Validate against production data
- [x] Generate keycount for every node, for every prefix

[carver]

The keyspace is too large to have an entry for every content ID, of course. So an area/bar chart will necessarily cover a range of content IDs for every bar. The edges of coverage, therefore, would only include part of some data points. I think the right approach here is to only increment the bar by 1 if a client claims to be interested in keys at both extremes of the bar's range. We probably want to under-count replication, when not exactly precise.

This has a couple awkward things to watch out for:

• Very small radius nodes will not cover the width of a single bar, so will not impact the replication. (we need to make sure the resolution is high enough, ie~ the width of the bars are small enough, to avoid losing too much information about these nodes)
• Because of the non-continuous nature of XOR distance, it's possible that a node covers the edges of a bar, but not some data in the middle. My intuition says this is a small enough problem that we can ignore it, but I am not sure I could convince a skeptic of that. The higher the resolution, the less of a problem that this should be.

So generally, we will want high resolution (narrow bars), as much as we can tolerate performance-wise.

[carver]

Thanks for hinting that fractional calculation of the keyspace is probably straightforward, @morph-dev . It is! I'll implement it that way from the start.

(I ended up writing a little python script to brute-force an 8-bit keyspace with a few different xor-distances, and make sure my closed-form solution was working correctly)

If you're interested, I'm happy to write it up or hop on a call. Otherwise, you'll probably see the prototype in a glados PR in the next couple of days.

[morph-dev]

Great! One other thing that crossed my mind is that while fractional calculation is good, we probably also want to know the number of nodes that fully cover the keyspace/domain. And we can show both information using stacked histogram.

[carver]

Yeah, I think that categorical split is helpful, I added it.

It will also probably be helpful to split by client (though I may hop onto something different, at the moment).