[rss] Cope with configuration mismatch

[smklein]

Note: This issue tracks follow-up work from #686

RSS configuration, as implemented, causes a number of datasets and services to be implemented on specific sleds. This configuration is read from a config file and transmitted to each sled.

On a "real rack", we expect the RSS to be driven with an operator-in-the-loop. Under testing conditions, however, it is convenient to automatically deploy this configuration throughout the rack. To cope with this, we:

• Parse a requested configuration file
• Apply the configuration to a number of sleds
• Record the configuration we just applied

This way, on subsequent boots of the system:

1. If the configuration was already applied, it does not need to be re-applied
2. If the configuration changes, we can detect this has happened

If we find the configuration mismatch (bullet 2), we currently throw an error, requesting a manual re-installation of the system. This was chosen for the sake of simplicity; we plausibly could cause the system to migrate from the old configuration to the new configuration by removing all Oxide datasets, and redeploying them all.

[andrewjstone]

No longer required as of #7185