Fix errors in cluster resource

[Bj0rnen]

• GET would respond with empty list on non-existing clusters. Now it returns NOT_FOUND as it should.
• Now total_segments in RepairRunStatus is correct.
• Default limit on number of repair runs to show for a cluster is changed from 10 to MAX_INT
• Order repairs for a cluster by end_time and then start_time, descending. Runs with no end_time list before anything else.

[Yarin78]

+1