Support API for stopping all running instances in a certain sled or all sleds

[askfongjojo]

Until sled-agent supports the mechanism of re-spinning propolis zones after they've been bounced or purged (omicron#3633) and live migration can be used in the field, stopping and starting VMs is the only way to evacuate VMs prior to sled SW/HW maintenance. Currently, neither Oxide technician nor operators can stop VMs on behalf of their owners. We'll need this ability to avoid telling users to force their VMs to failed state, then delete and create them again (from a customer pov, VMs are "broken" after software update).