HostToSp::HostBootFailure needs to be accessible beyond humility

[rmustacc]

There's a TODO in handling of various host failures that says to forward to MGS. Whether that, an ereport, or something else, we should make something discoverable as we cannot get access to this in an in-rack context and don't always have the luxury of hoping that we get another lap to confirm the information about what didn't happen.

This also applies to the panic and similar cases.

[hawkw]

I opened #2140 for generating an ereport for host panics and boot failures. I think we will also need to add something in the control-plane-agent protocol for getting the host panic/boot failure data out though, as the buffers for those are too large to fit in an ereport.

I think the approach here will have to be one where the ereport informs upstack software about the boot failure/panic, but a separate API is used to actually read the data out. That will probably also make it easier to consume it via faux-mgs in bench and test use cases.

[hawkw]

This is somewhere on my todo list (and would be happy to bump it up if it's become urgent). But, if anyone else gets around to it before me, please feel free to ping me if you have any questions about the ereport bit!