Fix testing of auto checkpointing

[vasslitvinov]

This PR fixes the shortcoming of this idiom in tests/optioned-server/auto-checkpoints.py:

assert directory_exists_delayed(autockptDir, 10)
delete_directory(autockptDir)

where the second line is executed after auto-checkpointing has started and therefore the directory has been created, and before auto-checkpointing is completed.

In such a case, removing the directory causes auto-checkpointing to fail. This goes unnoticed because the client has no way to know about it, however it throws off the carefully-carved sequence of events that the test code is intended to cause / observe, and testing fails.

How to detect when the server has finished creating a checkpoint, which we need before proceeding with next steps in testing? I added:
ak.client.wait_for_async_activity()
that makes the server wait for an in-progress auto-checkpointing, if any, to complete.

An alternative is for the server to create a marker file in the checkpoint directory upon the completion of a checkpoint. I did not implement this, however we may still want to add this. Another way is to "tail" arkouda.log to see if it ends with a line containing "finished autoCheckpoint".

Incidental changes:

• Add currentTime() as a proposed replacement for Time.timeSinceEpoch().totalSeconds().
• Prevent interruption of the server's "idle time" by a couple of messages like ruok and wait_for_async_activity.
• Skip the processing of a message on the server if there is an error early in message handling.

[ajpotts]

Looks good!

[vasslitvinov]

@ajpotts I just rebased my branch down to a single commit on top of the current master, while removing an unintended change to Logging.chpl - which motivated the rebase. I will keep an eye on the CI checks.