Semaphore `find_instance` blocks for up to 2 min

[icristescu]

Running a bootstrapped tezos node (1 block per minute) sometimes emits warnings

+306527809120us index_unix [WARNING] Semaphore find_instance was blocked for 47.988s
...
+211841313500us index_unix [WARNING] Semaphore find_instance was blocked for 2min33s

We should investigate - for instance we can report in the merge logs the time the t.rename_lock was taken.

[craigfe]

My WIP catapult tracing that was able to show this problem in combination with the bootstrap trace: https://github.com/CraigFe/index/tree/tracing. CC: @Ngoguey42.