want SP restart IDs in MGS messages

I left a note about this in https://github.com/oxidecomputer/hubris/issues/2212#issuecomment-3287413613, but it deserves its own issue. The control plane's fault management uniquely identifies a restart of a service processor via a 128-bit restart ID, which is generated by the RNG driver and posted to packrat, here:
https://github.com/oxidecomputer/hubris/blob/3b9b40d840984e3cca6d40e73f020f96a40538c5/task/packrat/src/main.rs#L507-L514

The restart ID is then included in every ereport received by the SP.

I'd really like these same restart IDs to be included in messages sent through the control-plane-agent protocol; in particular, the messages that drive SP sensor metrics in Oximeter (collected by MGS) and the messages that drive SP inventory collection in Nexus. This way, we can associate that data with a restart ID that produced a particular ereport.

In an ideal world, we would just include the restart ID in every single response from the SP. But, this might be painful, both in terms of eating up >16 bytes of the UDP packet and also that it would be a MGS API breaking change, which are always unpleasant. We could also add a new MGS request to read the restart ID. This would require code in upstack software that makes multiple requests to the SP, such as the MGS metrics code, to read the restart ID both before and after reading other data (in this case, the sensor readings), to ensure that the SP did not restart at some point during the sequence of multiple requests for other data. If it has, it wouldn't be possible to determine at what point exactly it restarted, so the control plane would have to throw out everything and start over from the beginning.