Fixup jasmine2 to have more consistent CI runs

[archmoj]

Recently the Jasmine2 container started to constantly failling on the CircleCI. It is also take a long time to fail i.e. about 15-20min.
This PR rearranged some of the tests in the gl2d_double_click_test simply by moving shared functions to the top, avoid nested describe blocks.
A new flag titled dbgl is added for a new container to isolate flaky gl tests.
Parallelism of test-jasmine2 increased from 2 to 3 so that the tests complete around the same time i.e. close to 5 min.
Finally the shard limit reduced to 1 in order to help reduce the chance of possible side effects of running one test on the other.

@plotly/plotly_js

[etpinard]

Thanks for taking a look into this @archmoj

You're right that test-jasmine2 has been very bad lately (after since we merged #3810)

Now, I can't say I'm a fan of this solution. I don't think we should be adding another container (that has a cost 💰 you know) and I don't think we should be adding another @tag (which has a maintenance cost).

Here are some ideas:

• Have you tried splitting up the @gl test into more _test.js files?
• There must be something in our @gl that's causing the flakiness? We should try:
  • Adding longer delays between drag commands
  • Adding delays between each it - to allow the container to (hopefully) slow down its CPU usage

[archmoj]

• Have you tried splitting up the @gl test into more _test.js files?

It is a very good idea. Namely the gl2d_double_click_test.js.

[archmoj]

Closing in the favour of #3859.