Phase out d3

[knolleary]

Node-RED has used D3 since day one of its existence. The original structure of the view code was based on a D3 example (can't track down the actual one, but was similar to https://observablehq.com/@d3/force-directed-graph).

Whilst it has remained quite fundamental to how view.js works, there is something to be said for phasing out its use. There are optimisations we can make in the code if we aren't having to work in the d3 way.

We're also on an ancient version of d3. Faced with the choice of phasing out its use, or upgrading across multiple major version changes, I want to explore the former.

This PR is a work-in-progress towards that goal.

There are a handful of places where we use d3 for no reason other than it was convenient at the time - and vanilla javascript is more than good enough. That's where I'll start.

[coveralls]

Coverage remained the same at 68.193% when pulling 11a6925 on phase-out-d3 into 30ea300 on dev.

[knolleary]

Current occurrences of d3.

• js/settings.js - 1 reference - logging d3.version - trivial to remove once we get rid of the rest
• js/ui/palette.js - 1 reference. Part of link splicing - uses the data() function map the hovered dom element back to the actual link data element. Once we figure out how we want to replace d3 in view.js this functionality will need to be taken into account.
• js/ui/view-navigator.js - 7 references. Uses the select/enter/exit functionality to generate the mini-map view.
• js/ui/view.js - 253 references - ha ha ha. That's going to be some work.