Three again

[Fil]

Fixes:

• [x] #152 & #113 upgrades forceRadial to offer a preferred angle, and use accessors for x, y, angle
• [x] #81 gentler forceCenter, with a strength
• [x] #17 Add tests

Supersedes #172 #173 ; because of #113 it is a breaking change.

[Fil]

Requesting a review from @enjalot

Demo of forceRadial.angle at https://observablehq.com/d/9187c989f882beee

[enjalot]

I took a closer look at forceRadial and don't see anything problematic. I made a deterministic fork of your example to help me pull apart the demo: https://observablehq.com/d/8f912e218dd99d61

It would be interesting to plot the non-force radial layout as way of understanding (visually verifying) the underlying data distribution. I played towards this here: https://observablehq.com/@enjalot/triangle-world?collection=@enjalot/triangle-world so that I could demonstrate how well UMAP was doing conceptually here: https://observablehq.com/@enjalot/triangle-world-umap?collection=@enjalot/triangle-world

I'd like to do one for forceCenter in this way as well, if only to give people a better sense of what "softer" means.

[Fil]

Closing this PR since:

• forceRadial.angle is abandoned for now (#152)
• x and y accessors are added in #176 (#113)
• tests are added in #177 (#17)
• forceCenter.strength was added by https://github.com/d3/d3-force/commit/c46be8de85b5d69cbe0ae426cf36fadeec9e2de7 (#81)