Increased WebGL 2D projection precision - updated dependencies and mock

[monfera]

The preceding changes (#1033) vastly improved the precision of projections but further testing found realistic scenarios where the resolution was subpar. This PR is just an explicit dependency version bump as plotly.js itself didn't have to change. The nature of the changes in the WebGL dependencies was the switch of scaling and translation, an example of which is: https://github.com/gl-vis/gl-scatter2d-fancy/pull/3/files

The image mock also changed, as apparently, the WebGL images generated in the test environment aren't quite faithful to regular operations, and a dependency precision change altered one of the traces.

[monfera]

@etpinard Some tests on codepen, with powers of 2 steps, starting with 1ms and ending in the iron and bronze age, respectively:

• non-fancy: http://codepen.io/monfera/full/YpPwNN
• fancy: http://codepen.io/monfera/full/VmYeQb/

[etpinard]

@monfera the change seems sound to me.

Can you explain why your patches generated an image diff:

Looks like the rightmost marker pt isn't showing anymore. Is that expected?

[monfera]

@etpinard Yes I initially suspected that it was a clipping issue, in that maybe the projection didn't work with the point size, i.e. it thought that the point is invisible. But it's not the case. It's a similar artifact to the line artifact - if you run it interactively e.g. by putting the mock into a jasmine test, it's good.

[etpinard]

Interesting.

if you run it interactively e.g. by putting the mock into a jasmine test, it's good.

So you're suspecting discrepancies in our nw.js image server? If so, we should try to find a way to test the increased precision in our jasmine suite.

[etpinard]

💃

[etpinard]

Thanks @monfera