Reorder CI jobs to allow for more concurrent PRs

[mvines]

• move goes to secondary, because we're not actively doing much with it right now
• coverage must pass first, then stable, and then the PR is eligible for perf/bench/local-cluster. This will hopefully filter out red PRs and unlock machines faster

[danpaul000]

+1 to moving move.

Adding both the wait steps into the mix will increase the minimum amount of time a single PR takes to get merged, assuming no bottleneck, such as off-peak-hours commits for our devs in Asia. Does stable depend on coverage, or could those two be run in parallel?

[danpaul000]

I guess I'm just wondering where the most-likely to fail test step is and put that towards the front of the work queue. Do a lot of PRs fail on coverage? Let's put them through the most restrictive filter first.

[mvines]

Adding both the wait steps into the mix will increase the minimum amount of time a single PR takes to get merged, assuming no bottleneck, such as off-peak-hours commits for our devs in Asia.

Yeah I know :(

Does stable depend on coverage, or could those two be run in parallel?

Doesn't depend but a lot of the stable tests are also run in coverage, so if one of them fails then I don't want the other one to run at all, and coverage is faster so that was my logic for this ordering