Update git submodules

* Update plugins/replication from branch 'stable-3.5'
  to c6665b860961603b38b67f28e9d84b60ccf70d50
  - Merge branch 'stable-3.4' into 'stable-3.5'
    
    * stable-3.4:
      Do not retry replication when local repository not found
      Ensure states are updated for canceled replication tasks
    
    Change-Id: Ic0e523c7c92222b49916aed7067836d4e6d2f6e7
    
  - Demote Files.list() errors when distributor is enabled
    
    Since Files.list() errors for the 'waiting' and 'running' dirs is common
    and normal when running with a multi-primary setup, demote them to
    "atFine()" level when the distributor is enabled. This avoids making the
    logs extrememly noisy during normal multi-primary operation.
    
    Change-Id: Iefc96fb79754dad9201afbc194d85194ef257198
    
  - Merge branch 'stable-3.3' into stable-3.4
    
    * origin/stable-3.3:
      Do not retry replication when local repository not found
    
    Change-Id: I6a8d0650ca24a4aac86fdafc819b028cf8864332
    
  - Do not retry replication when local repository not found
    
    When many repositories get created and deleted in a quick succession it
    may happen that a repository gets deleted before its replication task
    starts. Such replication task will keep retrying, possibly indefinitely,
    but has no chance to succeed.
    
    Another scenario where this issue can occur is when a repository gets
    created but a replica is not reachable for some time. If the repository
    gets deleted before the replica gets reachable again, the replication
    task will keep retrying but the local repository will not exist.
    
    When handling the RepositoryNotFoundException in PushOne, set the
    retrying flag to false. This ensures that this replication task is
    not retried and gets finished.
    
    Bug: Issue 15804
    Change-Id: Ia55c5ec1c961f4c2aec9ecee8056f22b436e9fda
    
  - Ensure states are updated for canceled replication tasks
    
    When using 'replication start --wait' we need
    ReplicationState.waitForReplication() to return when the task we're
    waiting on has been canceled, either through an admin action or because
    the replication distributor determined another node already completed
    it.
    
    Add a couple tests for PushAll that confirm this behavior was previously
    broken and is fixed now.
    
    Change-Id: I36320ae079af5d7673e05d20ddc94b42a9b04347
    
diff --git a/plugins/replication b/plugins/replication
index 20f1c21..c6665b8 160000
--- a/plugins/replication
+++ b/plugins/replication
@@ -1 +1 @@
-Subproject commit 20f1c213bc2c56bddce03cf8d4b0b63bd609fc20
+Subproject commit c6665b860961603b38b67f28e9d84b60ccf70d50