Add sled-identifiers to instance vCPU stats

[bnaecker]

This adds the sled UUID and baseboard info into the existing vCPU stats. This is an incompatible change, and will require dropping old data from the previous versions of the timeseries. That's unused today, and it's not yet clear how we'll actually manage the timeseries data across a schema change, so I'd like to include this now.

[bnaecker]

Here's a bit more context on the plan here. This adds some sled identifiers to the instance vCPU data, which is one piece of #5267. We don't yet have all the machinery in place for actually doing a timeseries upgrade, since I'm really just in the process of moving things to the TOML format that we'll used to describe those upgrades. But how we handle data across them is not yet clear.

In the meantime, I've also added the ability to drop individual timeseries by name in #6040. I'm planning to:

• Make this upgrade here, which is incompatible
• Update Propolis to use this new definition, populating the sled-identifiers in it using the interface I added in Add sled-agent endpoint for fetching sled identifiers #6086.
• Pull that updated Propolis into Omicron, and at the same time include a ClickHouse schema upgrade which does nothing other than drop the original virtual_machine:* timeseries.

[hawkw]

This seems reasonable to me, I think.

[bnaecker]

Thanks for the gut-check @hawkw