[Taskcluster] run Chrome and Firefox beta daily

[foolip]

To mitigate #19206.

[foolip]

There are downsides to this:

• More resources used, and the beta results aren't part of anyone's workflow (every time I try I run into a lack of runs)
• More work in a single task group, currently if any task fails none of the runs will end up on wpt.fyi

[jgraham]

I don't mind landing this, but if the beta runs are failing as often as not it seems like some work to address underlying causes (e.g. making us retry network errors) would be worthwhile.

In particular the downside of making stable runs more vunerable to failure seems worrying. I wonder if we should reduce the number of chunks also, now that we have better reliability. That's less overhead and less chance of network errors affecting a subset of chunks.

[foolip]

Yeah, I think improving robustness in the face of network errors would be good and likely would get us beta runs on most weeks. At this point I don't even know what the reliability is, exactly how to set up #14210 isn't clear. If we knew how often each type of run failed and the reasons, it would be easier to make a judgement.

I'll close this in favor of #14210 for now.