Real time metrics

[na--]

This fixes #614 and fixes (or at least helps with) #670

The refactoring is complete and I've fixed all of the old unit tests. I don't expect any issues, but I still want to write a few new tests that specifically check certain things in the new workflow, like metrics from setup() and teardown() and the real-timeliness of the metrics pipeline. Still, this this should be stable enough for code review and testing, especially to check if there are improvements to the aggregation characteristics.

[CLAassistant]

All committers have signed the CLA.

[codecov-io]

Codecov Report

Merging #678 into master will decrease coverage by 0.16%.
The diff coverage is 70.58%.

@@            Coverage Diff             @@
##           master     #678      +/-   ##
==========================================
- Coverage   64.83%   64.67%   -0.17%     
==========================================
  Files         101      101              
  Lines        8221     8229       +8     
==========================================
- Hits         5330     5322       -8     
- Misses       2549     2565      +16     
  Partials      342      342

Impacted Files	Coverage Δ
lib/runner.go	0% <0%> (ø)	⬆️
stats/stats.go	57.34% <0%> (-2.56%)	⬇️
lib/options.go	92.3% <0%> (-1.45%)	⬇️
js/modules/k6/metrics/metrics.go	93.33% <100%> (-0.42%)	⬇️
cmd/options.go	63.1% <100%> (+0.36%)	⬆️
js/modules/k6/k6.go	87.95% <100%> (-0.94%)	⬇️
js/modules/k6/ws/ws.go	71.42% <100%> (-1.12%)	⬇️
api/v1/setup_teardown_routes.go	55.26% <40%> (-0.3%)	⬇️
js/modules/k6/http/http_request.go	80% <81.81%> (-0.19%)	⬇️
core/local/local.go	79.1% <82.05%> (+1.1%)	⬆️
... and 3 more

---

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 cd557cf...a9a134a. Read the comment docs.

[luizbafilho]

LGTM

[na--]

I consider this fully done for now, even though we probably should refactor things again when we implement the arrival-rate based executor or the different rampup/rampdown methods or the native k6 clustering 😄. I did some simple performance tests yesterday and did not notice any regressions (it seemed like there was even a very slight performance improvement), so this should work for now and allow us to have better aggregation until we improve the whole engine/executor/"runner" architecture.