Move postsubmit build scheduling from LUCI to cocoon

[godofredoc]

This will allow to remove the current 45 minutes propagation time.

Migration

• flutter/cocoon
• flutter/packages
• flutter/plugins
• flutter/engine ci.yaml roller
• flutter/engine bringup targets
• flutter/engine - [ci.yaml] Migrate remaining targets to cocoon scheduler engine#32018
• flutter/flutter bringup targets: Migrate bringup post-submit targets to cocoon scheduler #107483
• flutter/flutter - host only target: Migrate host only targets to cocoon scheduler #107673
• flutter/flutter - linux/android: Migrate Linux android targets to cocoon scheduler #107852
• flutter/flutter - windows/android+ mac/android: Migrate Mac/Windows+android targets to cocoon scheduler #107737
• flutter/fluter - mac/ios - Migrate Mac_ios targets to cocoon scheduler #107944
• schedule commit targets in batch requests - Send batch requests when scheduling new commit builds cocoon#1999
• Schedule failed builds with higher priority - Schedule earlier failed builds with higher priority to open tree faster #108183
• Batch schedule samsung & mac host only tasks - Batch schedule samsung and Mac host only tasks via cocoon #108317
• deprecate /api/refresh-chromebot-status - Remove cronjob relying on /api/refresh-chromebot-status cocoon#1998, Remove refresh chromebot status API cocoon#2006
• update CI_YAML.md to deprecate luci scheduler: CI_YAML doc needs update to deprecate luci scheduler #109712

Features

• Rerunning Cocoon scheduled builds - [scheduler] Migrate rerun logic to Cocoon scheduler cocoon#1610
• Auto-retries for Cocoon scheduled builds - [scheduler] Add auto-retries on pubsub notifications cocoon#1657
• Batch on device builds - [scheduler] Support batch scheduling #100793

[CaseyHillers]

This is blocked on the scheduler supporting postsubmit triggers (tracked in #81795). Once that is complete, there will be a migration plan to move tests from LUCI to Cocoon.

[CaseyHillers]

The migration ran into a blocker as a few DeviceLab tasks checkout Cocoon as soon as the task starts, and flutter/flutter takes ~30 seconds to mirror. This was causing #100866

I have a few ideas:

1. Move the dependencies downloads and checks to being the first step (this unblocks the current steps, but causes race conditions down the line)
2. Add longer retries to the source checkouts (a minute would suffice). The engine gclient can retry up to 2 hours
3. Separate the scheduling logic to be triggered on the git-on-borg mirror (more complicated)

[keyonghan]

Blocker #107599 before migrating non-bringup targets.

[keyonghan]

Now everything is in cocoon.

[github-actions]

This thread has been automatically locked since there has not been any recent activity after it was closed. If you are still experiencing a similar issue, please open a new bug, including the output of flutter doctor -v and a minimal reproduction of the issue.