This PR adds the new "backup" command to redis-cli's cluster manager.
As its name suggests, it basically saves a backup of the cluster to a local directory.

The backup can be launched with:

redis-cli --cluster backup NODE_IP:NODE_PORT BACKUP_DIRECTORY

where NODE_IP:NODE_PORT is the entry point address of the cluster (as usual).

ie.

redis-cli --cluster backup 127.0.0.1:7000 /tmp/mycluster_backup

The backup will consist of:

• An .rdb file for each master node of the cluster
• A JSON file (nodes.json) containing the cluster configuration

The nodes.json file will contain something like this:

[
  {
    "name": "1cb877c71b1c27aa06a65cb987a27ddbf92b09e7",
    "host": "127.0.0.1",
    "port": 7000,
    "replicate": null,
    "slots": [[0,1],[3,5460]],
    "slots_count": 5460,
    "flags": "master",
    "current_epoch": 8
  },
  {
    "name": "20ad4ede36e5ac6a18a3c0efd673c37f34dfc834",
    "host": "127.0.0.1",
    "port": 7002,
    "replicate": null,
    "slots": [[5461,10922]],
    "slots_count": 5462,
    "flags": "master",
    "current_epoch": 3
  },
  {
    "name": "e7a68de2e71595c8fa4c2e84dc889c16bb558c8c",
    "host": "127.0.0.1",
    "port": 7004,
    "replicate": "20ad4ede36e5ac6a18a3c0efd673c37f34dfc834",
    "slots": [],
    "slots_count": 0,
    "flags": "slave",
    "current_epoch": 5
  },
  ...
]

The command will perform an preliminary check on the cluster. If the cluster has some problem,
the backup will be saved anyway, but it will display a warning message and the number of errors found will be saved inside every node in nodes.json (ie. "cluster_errors": 2).

In the next days I'll add more info to nodes.json (ie. migrating/importing slots), and in the near future I'll also implement a "restore" command. Anyway, at the moment, the backup can be manually restored by using the saved rdb files and info taken from the node.json.