Expose results for experimental browsers in a way that is easily linked

[foolip]

This is the frontend part of web-platform-tests/results-collection#146.

@lukebjerring has suggested we could use a label system to be able to mix-and-match. Anything that will allow us to have one link for Chrome+Edge+Firefox+Safari stable, and another for their most recent pre-stable builds, would do the job.

[foolip]

Issues for getting experimental results:

• Chrome + Firefox on Taskcluster
• Chrome Dev
• Edge Insider Preview
• Firefox Nightly
• Safari Technology Preview

[jugglinmike]

Results for experimental builds of Chrome and Firefox are now available in the same format/location as the results for stable builds.

[foolip]

@jugglinmike, are the unsplit wptreport JSON files also preserved somewhere?

[jugglinmike]

Nope. We can start doing this if you like, but storing the complete WPT report is part of @Hexcles' proposal for the results receiver. Unless there is some specific value in the unsplit files themselves (and depending on how much time Robert will need), it might be wise to wait for the new results receiver.

[foolip]

Nah, if you don't happen to have it then let's treat it like all other results and transition by re-merging them.

[lukebjerring]

I'm going to do this in a few steps.
First up, I'll add ?label=unstable but spoof it under the hood to change the browser names to [name]-experimental during the query, and we'll set that sit in staging until results-receiver progresses.

[foolip]

@Orphis asked me about experimental runs on wpt.fyi today and I pointed him at https://wpt.fyi/results/?label=experimental.

@lukebjerring, is all that remains of this issue to make experimental runs more discoverable?

[Hexcles]

This falls under the umbrella issue I just filed #379 . Perhaps we should create a project.

[foolip]

@TimothyGu has been using wpt.fyi for his work on HTML document.open() interop, and told me that finding this label was difficult. So bump :)

[lukebjerring]

Closing this as a dupe of #379, which will be fixed via #535

[foolip]

There is something left beyond #535, which is to make it easy to discover that ?label=experimental is a useful view, which isn't quite the same as the ability to pick an arbitrary set of runs.