[BP-1.5][FLINK-10309][rest] Before shutting down cluster, wait for asynchronous operations

[GJL]

What is the purpose of the change

Wait for the result of asynchronous operations to be served before shutting down the cluster. This is necessary for the "cancel with savepoint" operation. If we do not wait for the result to be accessed by the client, we may shutdown the cluster, and the client gets a ConnectionException.

cc: @zentol

Brief change log

• Before shutting down cluster, wait for asynchronous operations.
• Log stacktrace if checkpoint cannot be ack'ed.

Verifying this change

This change added tests and can be verified as follows:

• Added test to RestServerEndpointITCase to verify that handlers are closed first.
• Added unit tests for CompletedOperationCache.
• Verified the changes by submitting and cancelling with savepoint of a job in a loop.

Does this pull request potentially affect one of the following parts:

• Dependencies (does it add or upgrade a dependency): (yes / no)
• The public API, i.e., is any changed class annotated with @Public(Evolving): (yes / no)
• The serializers: (yes / no / don't know)
• The runtime per-record code paths (performance sensitive): (yes / no / don't know)
• Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Yarn/Mesos, ZooKeeper: (yes / no / don't know)
• The S3 file system connector: (yes / no / don't know)

Documentation

• Does this pull request introduce a new feature? (yes / no)
• If yes, how is the feature documented? (not applicable / docs / JavaDocs / not documented)