Secret panic hidden in `test_volume_replace_downstairs_then_takeover`

Looking at state transitions in test_volume_replace_downstairs_then_takeover, I see the following:

INFO [0]  New New New ds_transition to WaitActive, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95
INFO [1]  WaitActive New New ds_transition to WaitActive, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95
INFO [2]  WaitActive WaitActive New ds_transition to WaitActive, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95
INFO [0]  WaitActive WaitActive WaitActive ds_transition to WaitQuorum, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95
INFO [1]  WaitQuorum WaitActive WaitActive ds_transition to WaitQuorum, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95
INFO [2]  WaitQuorum WaitQuorum WaitActive ds_transition to WaitQuorum, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95

INFO [0]  Active Active Active ds_transition to Replacing, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95
    INFO [0]  New New New ds_transition to WaitActive, session_id: 638e7625-29e5-4a1f-b779-45f5fdbf1f17
    INFO [1]  WaitActive New New ds_transition to WaitActive, session_id: 638e7625-29e5-4a1f-b779-45f5fdbf1f17
    INFO [2]  WaitActive WaitActive New ds_transition to WaitActive, session_id: 638e7625-29e5-4a1f-b779-45f5fdbf1f17

INFO [1]  Replaced Active Active ds_transition to Disabled, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95
INFO [2]  Replaced Disconnected Active ds_transition to Disabled, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95

    INFO [0]  WaitActive WaitActive WaitActive ds_transition to WaitQuorum, session_id: 638e7625-29e5-4a1f-b779-45f5fdbf1f17
    INFO [1]  WaitQuorum WaitActive WaitActive ds_transition to WaitQuorum, session_id: 638e7625-29e5-4a1f-b779-45f5fdbf1f17
    INFO [2]  WaitQuorum WaitQuorum WaitActive ds_transition to WaitQuorum, session_id: 638e7625-29e5-4a1f-b779-45f5fdbf1f17

INFO [0]  Replaced Disconnected Disconnected ds_transition to WaitActive, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95
INFO [1]  Replaced Disconnected Disconnected ds_transition to WaitActive, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95
INFO [2]  Replaced WaitActive Disconnected ds_transition to WaitActive, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95

It turns out that going from Replaced to WaitActive is illegal, and we in fact see a panic in the logs:

Nov 03 17:51:48.435 INFO [0] 77d7dfc7-e462-496e-9be6-876f133cbcfe (c9f8ab22-b719-48a1-be58-eb7e8a733e95) Replaced Disconnected Disconnected ds_transition to WaitActive, session_id: c9f8ab22-b719-48a1-be58-eb7e8a733e95
thread 'test::integration_test_volume_replace_downstairs_then_takeover' panicked at '[0] 77d7dfc7-e462-496e-9be6-876f133cbcfe Initializing Negotiation failed, Replaced -> WaitActive', upstairs/src/lib.rs:6334:21

This happens to work, because the tokio::sync::Mutex<Downstairs> just silently unlocks on panics (and kills the task, which is unused in this test). However, it's causing difficulties in my experiments with std::sync::Mutex, because that flavor of mutex is poisoned on panic.

Start with the integration test test_volume_replace_downstairs_then_takeover and inspect the transition handling around upstairs/src/lib.rs:6334. Reproduce the test and trace the Replaced to WaitActive transition; done should address the reported panic rather than relying on Tokio mutex behavior.