feat(mu): add task recovery for pending tasks

[jfrain99]

Background

#1074 added an exponential backoff to the MU retry mechanism. The retry mechanism attempts to process a message, and on an error, it adds it back to the queue after a (now exponential) timeout. Because of this change, tasks can sit in these timeouts for up to 2 days. Before, they could only wait 5 seconds.

Problem

This greatly increases the chances of tasks being in a retry delay when a MU cycles. These messages are lost! We need to make it so that tasks waiting to be retried are not lost on a cycle / restart.

Proposed Solution

Currently, tasks are added to a sqlite table (tasks) and removed from the database when they are dequeued and begin processing. Instead, we can add a status column to the tasks table that marks whether a task has be dequeued. Then, when the message is successfully processed (or runs out of retries), the task is removed from the table. When recovering the table on cycle, all pending statuses should be overwritten.