Improvements

• Allow empty regions to be scheduled and use a separate tolerance configuration in scatter range scheduler #4117
• Improve the performance of synchronizing Region information between PDs #3933
• Support dynamically adjusting the retry limit of a store based on generated Operator #3744

Bug Fixes

• Fix the issue that PD does not fix the down peers in time #4077
• Fix the issue that the replica count of the default placement rules stays constant after replication.max-replicas is updated #3886
• Fix a bug that PD might panic when scaling out TiKV #3868
• Fix a bug that the hot region scheduler cannot work when the cluster has evict leader scheduler #3697