After unrecoverable errors, ZeebePartition accesses closed raft context

[oleschoenburg]

Describe the bug

When handling unrecoverable errors, ZeebePartition shuts down the raft context:

zeebe/broker/src/main/java/io/camunda/zeebe/broker/system/partitions/ZeebePartition.java

Line 377 in e15428b

context.getRaftPartition().stop();

which triggers an async transition to inactive:

zeebe/atomix/cluster/src/main/java/io/atomix/raft/impl/DefaultRaftServer.java

Line 135 in ae10d8f

context.transition(Role.INACTIVE);

Because this transition is async, the ZeebePartition continues to run on a closed raft context which can cause various issues, for example attempts to access a closed journal. This can then cause further issues, for example failed transitions.

Expected behavior

When the ZeebePartition decides to close the raft context, it should first remove itself as a role change listener, trigger the transition to INACTIVE itself and only close the raft context on completion of the transition.

related to #10776

[megglos]

Impact: as of now Log noise if a partition is dead, potential for critical bugs
Will put into ready to get it solved before it becomes an issue.