Context: A master PFQ build could end in several ways, including but not limited to:
1. master ran successfully, ran PublishUprevChanges, and uprev'd successfully
2. master ran successfully, but didn't actually have chrome revisions (takes about 30 minutes for slaves to report back), and didn't run PublishUprevChanges
3. master failed MasterSlaveSyncCompletion, ran PublishUprevChanges, but did not uprev because a slave failed

In the third case, the output from PublishUprevChanges and cros_mark_as_stable makes it seem like the uprev happened anyway.

For example, master-chromium-pfq build 3308 failed due to other builder failures[1]. So, as intended, it did not actually uprev, according to GoldenEye[2] (see chrome "55.0.2843.0", chrome os "8751.0.0-rc1", started "08-28 20:22").

However, the output from PublishUprevChanges is virtually indistinguishable from a build that *does* successfully uprev for realsies.

[1] https://uberchromegw.corp.google.com/i/chromeos/builders/master-chromium-pfq/builds/3308
[2] https://cros-goldeneye.corp.google.com/chromeos/console/listPfqBuild#/ (https://cros-goldeneye.corp.google.com/chromeos/console/listPfqBuild#/

Snippet from PublishUprevChanges step:

> peach_pit-LATEST_RELEASE_CHROME_BINHOST.conf: updating LATEST_RELEASE_CHROME_BINHOST
> 
> Change-Id: I0b6044d22b4c94b808c717606ca452369cb0fd52
> 
> 01:00:10: INFO: Successfully pushed /b/cbuild/internal_master/src/private-overlays/chromeos-partner-overlay to cros-internal merge_branch:refs/pfq/staging_pfq_branch_1031067
> 01:00:21: INFO: For /b/cbuild/internal_master/src/third_party/chromiumos-overlay, using description Marking set of ebuilds as stable
> 
> x86-generic-LATEST_RELEASE_CHROME_BINHOST.conf: updating LATEST_RELEASE_CHROME_BINHOST
> 
> Change-Id: Ia88b05bf895a43cf2852fe98fd30cadb5b712093
...
> Change-Id: Ifdd2895fc3e98986a02995d318d58ecdfcfbec03
> 
> Marking latest_release for chromium-source ebuild with version 55.0.2843.0 as stable.