Add graph db scheduling logic

Assuming we use model schema outlined in https://handbook.gitlab.com/handbook/engineering/architecture/design-documents/knowledge_graph/#scalability-and-high-availability then:

• add a new service which schedules re-indexing of the repository on its primary graph node (primary node could be just first replica for the repository), if there are not any replicas yet, find a graph node where the repository's graph DB should be created (e.g. a node with least existing DBs or with free space)
• creates a zoekt task "create graph" for the replica

Related POC !189941 (closed)