Add master/slave history (faliover) metric

[no1seman]

It wiil be great to add one simple metric that will return box.info.ro==false => 1 box.info.ro== true> 0 on each metrics scrape. This will give us an option to draw a graph of history of master switching over all instances of cluster and then we can easily filter instances by replicasets (may be for this only metric add replicaset alias label for Cartridge apps and replicaset UUID for non-Cartridge ones?).

[opomuc]

What if we had multiple changes inside scrape_interval? It will not affect graphs, but can influence an application

[no1seman]

Well it's not a problem to set a trigger on box.ctl.wait_ro()/box.ctl.wait_rw() but I have no any ideas of how to graph that info without dramatic coplexity. Have you?