Improve cloud aggregation for different URLs

[cuonglm]

For cloud output, we currently don't aggregate HTTP metrics with
different url tag, but with the same name tag.

This PR improves aggregation by setting url tag to be the same as name
tag. It's not ideal, but seems like the least bad option in term of UX.

Fixes #1166

[codecov-io]

Codecov Report

❗ No coverage uploaded for pull request base (master@78858ef). Click here to learn what that means.
The diff coverage is 100%.

@@            Coverage Diff            @@
##             master    #1220   +/-   ##
=========================================
  Coverage          ?   73.77%           
=========================================
  Files             ?      147           
  Lines             ?    10695           
  Branches          ?        0           
=========================================
  Hits              ?     7890           
  Misses            ?     2345           
  Partials          ?      460

Impacted Files	Coverage Δ
stats/cloud/collector.go	72.61% <100%> (ø)

---

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 78858ef...e3912bc. Read the comment docs.

[na--]

Besides the inline comment, you should also actually call sc.ReplaceURLWithName() (or however the method is named) somewhere in here: https://github.com/loadimpact/k6/blob/c2d73ba070b5223d60e784401e9bda8d634440fe/stats/cloud/collector.go#L271-L272

[na--]

LGTM, only squash the commits before merging

[mstoykov]

I would prefer if you either:

1. copy the whole sample but only if there is
   1.1. url
   1.2. name
   1.3. url != name
2. don't use the sample's tags but work around them somehow - probably the more perfomant variant but will take more code
3. shelf this until we have more optimized tags/ metric samples and we can fix it without introducing neither a race condition nor a big perf penalty

[na--]

@mstoykov, see my inline comment why this isn't problematic from a data race perspective. Rather, the problem is that if you have multiple outputs, the cloud output's url=name replacement could affect metric tags in the other outputs as well, depending on the order they are defined 😞 I'd be fine with that corner case bug, because:

1. it's not terrible
2. not a lot of people use multiple outputs
3. the alternative would involve CPU-heavy things (copying string maps around) in an already performance-sensitive area of code

So I'm for merging this as it is now, and properly fixing it when we implement the better cloud aggregation.

[na--]

Hmm although the performance overhead of copying the maps around wouldn't be too terrible, given the fact that we'd only do it if name != url...