Monitor and act on frozen compactions

Based off #7118 , the compaction Job code should asynchronously monitor a running compaction in a container to detect if it has frozen. This could be by polling the compaction for the current number of rows processed and if no change is seen in a configurable time window, assume the compaction is frozen.

Remedial action probably wants some further discussion, but we could do some combination of:
1) Send an administrative alert, produce AWS Metric for display on dashboard
2) Terminate the compaction container
3) Put compaction job back on SQS queue.