Figure out why a pad value is needed in geo range box bounds calculations

[etpinard]

Referencing commits from https://github.com/plotly/plotly.js/tree/try-bigger-clippad

---

From #2030 (comment)

If I do Plotly.relayout(gd,{coastlinewidth:10}) (still using geo_orthographic), occasionally (I really can't figure out when ) the viewport shrinks:

FYI I was able to reproduce the bug using a simple relayout call - see 3ad3d1e

The problem can be fixed by increasing the clipPad geo constant from 1e-3 to 1e-2. Now, why? Without a clipPad some clipped projections (e.g. orthographic), compute different bounds for different geo.projection.rotation values. But this shouldn't happen. Bounds should be a function of the lon/lat axis range only. This is possibly due to rounding errors in d3, but my investigation stopped there.

@alexcjohnson you can uncomment this line to draw the lon/lat range-box polygon on the map if you feel like investigating further.

Bumping clipPath to 1e-2 does affect the baselines - see 3cd0ca5, but not noticeably. The jasmine tests are passing fine.

There might be other situations where clipPath: 1e-2 will fail too. Maybe setting it to 0.1 would be safer. Oh well, I couldn't get geo to break with clipPath: 1e-2.

---

and @alexcjohnson 's followup #2030 (comment)

I'm uncomfortable with this solution. For example, what if you zoom in sufficiently that 0.1 (or even 0.01) is a noticeable fraction of the range? I think we need to step back and look into where this is coming from a little closer.

The bug exists on master (in fact it's worse, you just get the top or bottom half of the viewport rather than a centered half, see below) so lets ignore it for this PR and make a new bug report for it.

[etpinard]

Branch

https://github.com/plotly/plotly.js/compare/try-bigger-clippad

made an attempt at fixing this. It appeared to have "worked", but the reason why is still a little mysterious, so I'll keep the branch on origin, but I won't open a PR.

@archmoj if you notice weird things on relayout calls after drag interactions on geo subplots, it might be of interest to take a closer look at this issue.

[gvwilson]

Hi - this issue has been sitting for a while, so as part of our effort to tidy up our public repositories I'm going to close it. If it's still a concern, we'd be grateful if you could open a new issue (with a short reproducible example if appropriate) so that we can add it to our stack. Cheers - @gvwilson