Tracking issues for enhance replica / leader schedule

[zojw]

Current root balance schedule strategies are in a very early stage, this issue tracks the shortage or coming works for it.

(They may not take action in recent development iterations, but mark them here to help us remember them)

• transfer leader base on node's group-leader-count
• reallocate replica based on node's replica-count
• transfer leader & reallocate replica by QPS
• take care of disk capacity(should not reallocate replica to full node and balance full node data to others)
• combine transfer-leader and reallocate replica (e.g. the group is very worth to transfer leader out but couldn't find a suitable target, should be taken more consider in reallocate replica logic)
• take care of replica raft-log lag(make transfer leader decision to no-lag replica as possible)
• take care of in-progress change when making new balance & reallocate decisions
• schedule & reallocate speed control