Currently intent registration has the same blocking behavior that the RSTL-based concurrency management system had: for example when transitioning from secondary to primary we wait for all operations with Intent::BlockingWrite and Intent::Write to drain.

When operations aren't drained on time, we fassert, which is no different from what we did with the RSTL-based system. However, on fasserting in the RSTL-based system we dumped the lock manager's contents which told us which thread was holding which lock, which was useful for diagnosing deadlocks.

We should dump similar information out for the intent registration system.