Add disk_space_ok option to master_conditions and slave_conditions

Add enum option "disk_space_ok" to the master_conditions and slave_conditions-settings. When enabled, running out of disk space will cause [Master] or [Slave]-flags to be removed from the server, which affects query routing. "disk_space_ok" should be enabled by default in master_conditions. Slaves are already put into [Maintenance] by default, so the option does not need to be active in slave_conditions by default.

The purpose of this change is to stop writes to a master server that is out of disk space to prevent corruption. Affects any router that respects [Master]-flag (e.g. ReadWriteSplit). This feature only works if disk space checking is configured. If also `switchover_on_low_disk_space` is enabled, the master may be swapped with a slave.

Original description:
Customer currently has MaxScale in primary replica configuration. When disk fills to 100% on Primary there can be corruption. Customer already has Critical alerts when disk reaches 95% full. Customer would like to have Primary change to Read-Only when this threshold is reached. Currently, maxscale supports switchover on disk full (and maintenance on disk full for replicas), but Customer would like to leave the cluster in-tact and support Read-only transaction mode on Master node and continue to detect node failures (replicas) while in this state.

Customer admin attempted to put Primary into read_only mode, however maxscale recognizes this and changes it back. Also, locking tables for read only will not work for the customer since it hangs all other connections.

Summary: request for feature enhancement to add a "read_only_on_disk_full" option for Maxscale similar to switchover_on_disk_full?