Issue Summary

Cancelling repository migration to mesh doesn't cancel the migration job if Bitbucket node is crashed while a repository is in migrating state

This is reproducible on Data Center: Yes

Steps to Reproduce 

1. Initiate Repository migration to mesh
2. While the repository is in migrating state, kill (using SIGKILL and not SIGINT) the Primary Bitbucket node process
3. Monitor the Repository state, it will hung in migrating state forever
4. When we issue cancel call to Migration job with API /rest/api/latest/migration/mesh/<jobId>/cancel the repository state will change to CANCELING but will not cancel

Expected Results

The migration job gets cancelled.

Actual Results

The migration job state will change to CANCELING & remain in this state forever.

Workaround

The Migration job and Repository state has to be updated on database
Update Migration Job State

UPDATE bb_mesh_migration_job
set state = 3, end_timestamp = current_timestamp
where id = <Migration Job ID>;

Update Repository State

UPDATE bb_mesh_migration_queue
set state = 5
where  migration_job_id = <Migration Job ID> AND repository_id = <repo_id>;

Update bb_job table

UPADATE bb_job 
SET progress_message='The migration was canceled', state=700, end_timestamp = current_timestamp 
WHERE id = <Migration Job ID>