sled agent: Mechanics of applying host OS update

[smklein]

The Sled Agent needs to be capable of applying the host OS image to an m.2 when instructed by Nexus.

As mentioned in the 2/17 control plane huddle, the sled agent should write to the "other" m.2 than the one which was originally booted. @jclulow mentioned that pilot has a mechanism for parsing libdevinfo to make this determination, which we should probably steal here.

Anyway, here's what we need to do:

• Expose an API for nexus to call, to transfer a host OS image to the sled
• Observe libdevinfo to see both m.2s, and pick the one we did not boot from
• dd (or moral equivalent) to actually write the image out

[smklein]

Related to #2361 - that's the recovery equivalent of this issue.

[davepacheco]

I think based on context this is talking about storing new phase-2 images onto a sled (not phase-1 images).

[davepacheco]

The underlying need here has been resolved, largely by a combination of #8570 and #8540 (and the work leading up to those).