Fix table delete routing to region leaders

[jt2594838]

Problem

In a 3-node cluster, table-model DELETE may return success without deleting any data.

The first replica in a Region replica set is not always the current IoTConsensus leader. When a DELETE request is routed to a non-leader
replica, the operation can silently have no effect.

Changes

• Reorder Region replica lists in the ConfigNode response so the current leader is listed first.
• Reorder a copy of the replica list without modifying persisted replica placement.
• Add a 3-node table-model integration test that sends DELETE requests through each DataNode and verifies that the data is removed from
  every replica.

Validation

• PartitionInfoTest: 4 tests passed
• Spotless formatting check passed
• git diff --check passed

[CRZbulabula]

LGTM!