Servicenow-Observer to provide ability to discover relationships tables separately from resource discovery

At present when discovering resources from servicenow we specify the tables to collect data from and have to be explicit with the parent and child classes because whilst resources can be discovered from the parent class and it traverses the children, the final stage where relationships are discovered only discovers the relationship from the parent table to another parent table and doesn't discover anything within the discovered child tables.

ie  say we discover these classes :

cmdb_ci_service,cmdb_ci_datacenter,u_cmdb_service_component,u_cmdb_service_lane,cmdb_ci_appl,u_cmdb_service_application,u_cmdb_service_environment,cmdb_ci_server,cmdb_ci_unix_server,cmdb_ci_aix_server,cmdb_ci_linux_server,cmdb_ci_win_server,cmdb_ci_solaris_server,cmdb_ci_mainframe,cmdb_ci_mainframe_lpar,cmdb_ci_cluster,cmdb_ci_ip_network,cmdb_ci_ip_switch,cmdb_ci_firewall_network,cmdb_ci_ip_firewall,cmdb_ci_ip_router,cmdb_ci_database,cmdb_ci_db_instance,cmdb_ci_db_db2_instance,cmdb_ci_db_mssql_instance,cmdb_ci_db_ora_instance,cmdb_ci_lb,cmdb_ci_lb_network,cmdb_ci_lb_service,u_cmdb_ci_tandem,u_cmdb_ci_tandem_controller,u_prc

discovering these classes is necessary because when the observation moves to discovering the relationships in /api/now/table/cmdb_rel_ci then unless each class is specified it doesn't perform a parent/child discovery of the resource and find the upstream and downstream relationships.  

If we go with the parent class, then it will discover the children, but never tests the children's resources for relationships within /api/now/table/cmdb_rel_ci  so you end up with no relationships drawn for the children.

This has the effect of increasing discovery time and can take 5+ hours as it has to go over the same data and discover children it has already discovered.

So by adding in the ability to have a list of classes to discover for resources and separately a list of classes to discover for relationships it means I can have a cut down list of classes for the first part of the observation and when this is complete it will use the relationship discovery classes to find the remaining relationships and this can then include the child tables.