crucible: shouldn't start scrub before activating volume

[iximeow]

in block/crucible.rs we kick off a scrub of the volume. the scrub code asks a reasonable question: XXX Can we assert the volume is activated?

well. not really. propolis kicks off the scrub and then activates the volume a little later in Backend::start. it seems like the scrub works by the good graces of the 120s start delay that Propolis provides.

in practice there's no reason to have a propolis-server sitting ready but with backends not started for two minutes, so we don't hit this. but it's not really.... right.

[leftwo]

I was thinking, if the 605 work is going to be reading the volume, we probably want to have it also start that work after the volume is activated, so perhaps whatever mechanism 605 stuff uses, we can just move the scrubber to that point as well.