Test for sqlite corruption and dataloss

[markstory]

Within the GCP sandbox attempt to cause SQLite corruption or database loss. We're not confident in how stateful sets will preserve sqlite across restarts and scale operations. Attempt the following:

• restarting brokers
• pod moves, have k8s reschedule pods on new nodes.
• force taskbroker to panic!
• force taskbroker to oom

Ideally these scenarios do not result in the loss of a sqlite database.