CASSANDRA-21426: AssertionError in hasReplicaWithOngoingRepair when parallel_repair_count > 1

[patrickclee0207]

An assetion error is thrown from AutoRepairUtils.java when calling StorageService.instance.getTokenMetadata(). works fine when only 1 node is allowed parallel_repair_count: 1. When you increase this > 1 you start getting the assertion error

the change is to instead use StorageService.instance.getTokenMetadata().cachedOnlyTokenMap()

patch by Patrick Lee; reviewed by TBD for CASSANDRA-21426

using CCM and 9 nodes, with the auto-repair config

auto_repair:
  enabled: true
  repair_check_interval: 30s
  repair_task_min_duration: 5s
  repair_type_overrides:
    incremental:
      enabled: true
      min_repair_interval: 1m
  global_settings:
    parallel_repair_count: 3
    allow_parallel_replica_repair: false

you can reproduce and get the assertion error. running the same in CCM with the patched version validates that there are no assertion errors and there were 2 nodes running incremental repairs concurrently.