GitLab

This includes two fixes:
1) In OnStateChanged, check CanSyncFeatureStart instead of checking for
   absence of disable reasons. CanSyncFeatureStart also implies no
   disable reasons, but additionally checks that the syncing account is
   primary, meaning that Sync the *feature* can start, not just the
   transport layer.
2) In UpdateSettingsForMigration, when determining the UKM state, add a
   check for IsSyncFeatureEnabled. The reason is that
   GetChosenDataTypes returns "all types" by default, even if Sync is
   disabled, or only the transport layer is running. So the old code
   could result in some false positives.
   This change is analogous to https://crrev.com/c/1343369.

Bug: 907027
Change-Id: Ie5eaf261b17da072dcd2cb8a596a7342e2fc8192
Reviewed-on: https://chromium-review.googlesource.com/c/1346131Reviewed-by: Thomas Tangl <tangltom@chromium.org>
Commit-Queue: Marc Treib <treib@chromium.org>
Cr-Commit-Position: refs/heads/master@{#610038}