Restructure Backup tests

The `BackupCorrectness` workload has grown to do a lot of different things, and can be difficult to understand and combine with other workloads. In response, many other backup workloads have been created, but these repeat a lot of the same logic. I think that ideally we should have the backup and restore in completely separate workloads, since workloads can easily be combined in test files, so I propose we get rid of the workloads that do both backup and restore together. At the same time, a variety of new backup projects (incremental backup, backup to blob-store, fast-restore) have led to a lot of duplicate code across different workloads, so we should consolidate some of these workloads.

No files, tests, or entry points are named. Start by locating the BackupCorrectness and other backup workloads, then map their shared backup and restore logic. Done means the workloads are separated and duplicate logic is consolidated without changing the intended backup behavior.

Written by the indexing model from the issue text.