gl2d better ticks

[etpinard]

🆘 this is debatably a breaking change 🆘

In brief, this PR removes a block (see commit 3d67177) which dates back to the first gl2d implementation and recently highlighted in bbac99b (part of the gl2d annotation PR #1301). Back in #1301, I thought that this block was necessary to correctly draw the gl2d axis ticks, but I was wrong. That block was plain wrong from the start. In more technical terms, ax.setScale() sets the correct ax._length, no need to override it using glplot.viewBox.

Removing this block makes the gl2d baselines more on-par with gl2d. To demonstrate, commits d63578b and fbeb5e9 regenerate the scattergl baselines as SVG scatter traces, so comparing (in swipe mode especially) the baselines in commit 32efccd shows little difference between the SVG and gl2d versions. But comparing the new baselines with the old in the combined diff shows some pretty significant differences.

So, what do you think? Can this be considered a bug fix? Or, should we wait for v2?

cc @alexcjohnson @dfcreative @monfera @rreusser

TODO

• fixup jasmine tests which rely on wrong axis range values

[etpinard]

Full disclosure, I found out that the block removed in 3d67177 was wrong while working on scattergl lasso (PR #1657). As of the latest commit, 1D select boxes have the wrong width (because they have the wrong ax._length):

[monfera]

Better unity looks great and on first sight I don't see a negative with it, though my work on gl2d didn't touch on the axis ticks in this regard.

[alexcjohnson]

Looks great, I'm happy to call this unification a bug fix.