Heatmap equal gaps

[alexcjohnson]

Fixes #2142 - heatmaps with more than 3 bricks in one direction had nonuniform gaps. Looking into it in more detail, the bricks at the edges were also not centered on their desired data values, as the gap was applied only to the inside edges of these gaps.

This PR makes all the gaps symmetric (aside from rounding to keep the brick edges as crisp as possible).

Note the two commits here:

• 2802c79 just changes the mock & baseline image to make the issue obvious
• 0d1709b changes the algorithm and updates the baseline image and jasmine tests - please see the image diff in that commit to get a feel for what the new algorithm does differently. In particular, the bricks are all now a bit smaller than they were before, since some of the gaps had to expand to be as big as requested, as well as adding gaps around the edges to keep the bricks centered.

cc @CalvinFernandez @etpinard

[alexcjohnson]

We've seen this one fail a lot - and this morning seems like circle is super stressed or something, every intermittent test failure we've ever had seems to be biting us today. Choropleth needs to make a request for map data - it's local but still perhaps slow? Just give it some more time before we fail.

[etpinard]

it's local but still perhaps slow?

Actually, no. Geo subplots in jasmine tests get their map data from https://cdn.plot.ly/world_110m.json and friends.

You brought this up previously (see #163) but I never got around to fixing that.

---

👌 for bumping the timeout interval.

[etpinard]

Excellent fix & tests 💃