This change fixes 2 pathologies with pali chaos' stepdown lockout handlers:

• If a node restarts, it will reset its in-memory lockout, but pali chaos mirrors the timer without this logic. We fix this by resetting the mirrored blocker when we see the node restart.
• Even so, pali chaos may legitimately generate an interleaving which can not produce a leadership change due to the lockout. We fix this by not considering such an inability to be a test failure. Since pali chaos is a randomized fuzzer, we don't really lose coverage per-se, we just wait for it to produce interleavings that do work, as opposed to considering the invalid interleavings a test failure.