Feature: support broken disk check while starting and adding disk dynamically

[hycdong]

Feature Request

In current implementation, if a disk is broken, replica server will core, and server can't be started automically. It can be started after adding the broken disk path into black_list manually. If the disk repaired, we should stop replica server instance firstly, then remove the disk path from black_list, then restart the instance. This is not convenient when broken disk happened.

rdsn#834 supports disk check while initialization, if a disk is broken, its path will be excluded and replica instance can start succeed.
rdsn#839 supports a rpc for adding a new disk, no need to restart the replica instance.
NOTICE: if you add a totally new disk, not in the config files, you should add this disk path and tag in data_dir configuration.

[replication]
+ignore_broken_disk=true