Add Dropwizard Task to flush output and upload.

[prb]

This would be used when bringing down a node (after taking it out of the active pool) to ensure no data was lost.

[prb]

(From discussion with @lorrin.)
Task is one option here, and a CLI command is another. For the CLI command, the workflow would be:

1. Shutdown the service.
2. Invoke the CLI with a "flush logs" command that starts the Skid Road infrastructure and pushes out any "WRITING" or in-need-of-retry files.
3. Decommission the node.

[lorrin]

Initial command implementation in, handles happy-case where everything uploads cleanly on first try. But still needs:

• command-line argument to specify max wait time (currently waits indefinitely)
• more nuanced retry configuration so we retry more intensively than usual
• more aggressive retry that tries all items once rather than pausing after first failure

[prb]

It's OK if the container owns the status but not the retry execution itself. It can be up to the client of the command/task to check status.