Fix variable-looping-vus VU wobble during graceful ramp-down

[imiric]

This fixes the first bug mentioned in #1296, and adds a test that tries to ensure a smooth ramp-down.

I'm not proud of the test, and while it fairly consistently detected the bug and it might be useful for the second segment / execution plan issue, I'm concerned it might be flaky and would prefer a more deterministic approach of inspecting execution that doesn't rely on time.Sleep. Maybe inspecting metrics would be more reliable? Any ideas?

[codecov-io]

Codecov Report

Merging #1333 into new-schedulers will decrease coverage by 0.01%.
The diff coverage is 100%.

@@                Coverage Diff                 @@
##           new-schedulers    #1333      +/-   ##
==================================================
- Coverage           76.79%   76.78%   -0.02%     
==================================================
  Files                 160      160              
  Lines               12511    12511              
==================================================
- Hits                 9608     9606       -2     
- Misses               2394     2396       +2     
  Partials              509      509

Impacted Files	Coverage Δ
lib/executor/variable_looping_vus.go	92.62% <100%> (ø)	⬆️
lib/helpers.go	95.83% <0%> (-4.17%)	⬇️
lib/executor/vu_handle.go	94.28% <0%> (-2.86%)	⬇️
stats/cloud/collector.go	79.62% <0%> (+0.62%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9fb6466...2a7d7f9. Read the comment docs.

[na--]

LGTM. I'm not a huge fan of the test, but I'm not sure how to substantially improve it either...

The only improvement that I could think of is to involve the runner function in the test, instead of having it just sleep for one second. Maybe have a shared atomic that the runners increment and decrement, and check that at various intervals, seems like it might be a bit less susceptible and a more real way to test the actual execution. That said, the current test is probably ok, we can always investigate that if that test ever becomes flaky...

[imiric]

@na-- Good idea about involving the runner function. I'll consider that for upcoming tests, or if this one becomes an issue.