fix: don’t hold node lock if reboot is blocked

[jackfrancis]

Fixes #792

This PR moves the "is this node blocked for reboot?" foo earlier in the code flow in the main "does node need to be rebooted?" loop. The purpose is to determine whether or not the node is blocked for reboot prior to trying to acquire a node lock that will prevent another node in the cluster from being rebooted.

The purpose of blocking a node from rebooting is to implement node-specific safety checks prior to taking a node offline (an obvious side-effect of a reboot). These safety checks should be independent from the configuration that enforces how many nodes may go offline at any time (implemented via the node reboot concurrency config + the daemonset lock). This is what informs my thinking that we should simply check for a blocking condition prior to attempting to acquire a lock. If we are blocked for reboot, we should simply short-circuit the loop and try again later, like we do for other conditional outcomes that don't allow us to make forward progress.