need min-cluster-size option to avoid racey bootstrap

[Edward Hope-Morley]

Percona-cluster bootstrap is still racey in our charm when deploying > 1 unit. This appears to be a result multiple configure-and-restart actions as the service deploys. I frequently get a result whereby percona succeed at some point to bootstrap but then failed, causing the db to end read-only mode with 'wsrep_cluster_size 0'. The consequence is usually that deployment completes without errors until shared-db relations are added at which point I get an error like:

http://paste.ubuntu.com/11892179

Until we have leadership-election supported in the charms and stable in Juju we should add a config option similar to that used in the rabbitmq-server charm whereby the user can optionally tell the charm to wait for a certain number of units to be extant before it tries to configure/create the percona cluster.