Minimal Redis Sentinel support

[uglide]

For me the most important feature is that we can connect to a Sentinel, and we're using a Redis-Instance that is writable. Current situation:

Add all Redis-servers to RDM
Connect to first server
Try to add key, but get an error because the node is read-only
Connect to next server
Try to add key, this time it succeeds, because (by luck) we connected to the server that is currently the master.
Desired situation:

Add sentinel(s) to RDM (so instead of a Redis-Server entry, this would become a Sentinel-entry)
Connect to "sentinel", ask for R/W server
Try to add key
I've read the above, and while I agree that everything proposed above is a great addition to RDM, I also believe that maybe the core functionality (auto-select a R/W server using sentinel) would be the first step, and the other steps (GUI tree representation of sentinel and masters/slaves as well as GUI commands that trigger sentinel commands) would follow in later releases. But that's just my 2 cents :).

Full Sentinel support issue: #3399 (comment)

[obriat]

A support for role detection could be useful : indicate if the server is a Master, Slave, sentinel, if it's read only, if it's part of a shard or a cluster...

For example :

• if it's a sentinel, use a specific icon and display the elected Master : ROLE
• if it's a master, use a specific icon and show its slaves: INFO Replication.
• if it's a slave, use a specific icon and show its master: INFO Replication.
• A global check could be send to validate the consistency of the whole MASTER/SLAVE/SENTINEL setup.
• For sentinels: do not allow key display, but only console launch.

Bonus: allow a auto discovering of the configuration when adding a sentinel: add other sentinels, the master then the salves ?