Team building should take free disk space into account

If a FoundationDB cluster has a number of disks hit 95% disk space usage, then ratekeeper will kick in. It would be nice if one could then just add some storage servers to the cluster, and FDB would move data from full storage to empty storage. However, storage teams are built that mix the new hosts with old hosts, and thus make it much less likely that data movement will be able to choose empty teams to move data to. Thus, the current approach is instead to exclude all the existing storage servers and add enough new storage servers to fully contain the existing data.

It would be great if when adding N new, empty process that can form storage teams by themself, team building would first allow data to move from full to empty, and then worry about if it got team selection across all of the existing processes perfectly chosen.

The issue names no files, tests, or entry points. Begin by locating the team-building logic and related tests, then investigate how full and empty storage servers influence team selection; done means newly added empty servers can receive data from full servers before broader team-selection concerns are prioritized.

Written by the indexing model from the issue text.