Leader yield if any IO or disk down errors

[sanebay]

From @yamingk One issue I see is: when SM experienced disk down, if it happens on leader, HomeStore should yield leadership immediately, the reason is, if we rely on CM to do that, there needs to be some manual ops to kicks in or even if it is automated from CM layer, it also takes some time for CM to detect that. If Homestore can yield leadership it to other member, the other two copies can serve traffic immeidately without any SOPs need to be taken in production.