Add functional tests for task manager's functionality

[mikecote]

Our current testing of Task Manager relies on the tests of downstream services like alerting and actions. Now that we discovered a way to have better control on the server code when running, I wonder how many basic tests we can add that would ensure all of Task Manager's functionality works as expected.

Example:

• A recurring task gets rescheduled at X + interval after it runs successfully
• A recurring task that fails gets rescheduled at X + interval after it throws an error
• An ad-hoc task that fails gets re-attempted
• An ad-hoc task on it's last attempt gets deleted if it fails on the last run
• When an error is thrown during the polling cycle, Task Manager does X

The goal is to increase our level of confidence as we make future changes in Task Manager that we are not breaking functional requirements.

[elasticmachine]

Pinging @elastic/response-ops (Team:ResponseOps)