[Improvement] Slow LDAP scanner

Description:

LDAP import is really slow for large organizations

Steps to reproduce:

1. Run LDAP import with:

• Around 5000 users
• Around 400 groups

Due to LDAP_Sync_User_Data_Groups_Filter, Rocket.chat importer is doing per user:

• A LDAP request per entry in LDAP_Sync_User_Data_GroupsMap
• A LDAP request per entry in LDAP_Sync_User_Data_Groups_AutoChannelsMap

This is needed for nested groups in Active Directory I guess.

BTW, it is slow. Should be cool to be able to set an empty LDAP_Sync_User_Data_Groups_Filter.

If LDAP_Sync_User_Data_Groups_Filter is empty, Rocket.chat could then:

• On sync start, build a map with all users per groups in LDAP_Sync_User_Data_GroupsMap/LDAP_Sync_User_Data_Groups_AutoChannelsMap
• For every user, mapLdapGroupsToUserRoles() and mapLDAPGroupsToChannels() should check against this map in memory

Expected behavior:

Around 20 minutes scanning

Actual behavior:

Around 2 hours scanning

Server Setup Information:

• Version of Rocket.Chat Server: 3.7.1
• Operating System: Debian 10
• Deployment Method: tar
• Number of Running Instances: 1
• DB Replicaset Oplog: ??
• NodeJS Version: 12.18.1-1nodesource1
• MongoDB Version: 4.2.8

Client Setup Information

• Desktop App or Browser Version:
• Operating System:

Additional context

Relevant logs:

Start with LDAP_Sync_User_Data_Groups_Filter, LDAP_Sync_User_Data_GroupsMap, and LDAP_Sync_User_Data_Groups_AutoChannelsMap, then trace mapLdapGroupsToUserRoles() and mapLDAPGroupsToChannels() in the LDAP importer. Compare the current per-user requests with the proposed in-memory group map, and verify that large imports approach the stated 20-minute target without breaking nested-group handling.