Fix HTTP 415 when retrieving tags after nodes role switch

[outcatcher]

Make provider use master node instead of node with ID _node0 for getting tags

[outcatcher]

Possible fix of #562

HTTP 415 is not reproduced any more, though correct work of provider with switched RDS nodes requires setting

  lifecycle {
    ignore_changes = [
      "db",
      "availability_zone",
    ]
  }

So terraform won't treat switching nodes as a change in state ― availability zones got reordered, and change in AZ forces re-creation of RDS instance by default.