improvement(lb): add add_node_list.sh to add nodes with copy_pri after all copy_sec done

[foreverneverer]

What problem does this PR solve?

When we add new node for cluster, the copying operation will let the new node has very high load. And then the new node almost is unavailable.

What is changed and how it works?

This pr change the default order of copying primary preferentially to copying secondary preferentially by setting only_move_primary = true when we add new node and recover it after completed.

Before the total replica count is balanced, the new node won't have any primary until copying secondary completes. Because the new node won't serve user-requests, reads will not be influenced by the data-migration of replicas.

Check List

Tests (for the pr refactor the offline_node_list.sh, so need 2 manual test)

• Manual test for pegasus_online_node_list.sh (add detailed scripts or steps below)

  • deploy cluster with minos
  • add new node for the cluster using the script
  • check the nodes using nodes -d and the falcon
  • the result is: the cluster copy secondary preferentially and the falcon show the new node don't serve for read:
  • default version(the follow show the P999 latency):
    
  • new version:
    

• Manual test for pegasus_offline_node_list.sh(add detailed scripts or steps below)

  • deploy cluster with minos and show the running replica server node = 5
  • execute the pegasus_offline_node_list.sh and show the running replica server node = 4

Related changes

• Need to cherry-pick to the release branch
• Need to update the documentation
• Need to be included in the release note