Once we have smarter scheduling algorithms available in the taskgraph, we should have a try selector that uses them. Possibly it can be as simple as "run the exact same thing that we would run on autoland". It should also be made the default selector, so simply running mach try will invoke it.

Eventually we may want to implement the ability to supplement / prune the task list, but that can be implemented later on.

This allows us to change the default optimization strategy used in try pushes.
While probably not super useful to developers, it can help us easily test
changes to new and experimental optimizations on try.

This also changes the default to the 'bugbug_push_schedules' strategy, since
SETA is more or less random and shouldn't be used by 'mach try auto'. In the
future, we'll switch this back to simply using the default optimization as the
default will ideally be the best one that we have.

Depends on D60184

Neither of these selectors involve the user choosing tasks, so showing
estimates or saving them in the 'mach try again' history doesn't make much
sense.

Depends on D65746

https://hg.mozilla.org/mozilla-central/rev/12d3a008b699
https://hg.mozilla.org/mozilla-central/rev/ec855966b19b
https://hg.mozilla.org/mozilla-central/rev/3d0e23deb7f6