Fixes
- Fix deleting hosted data at the end of a contract blocking downloading
Fix volume usage accounting going negative
Fixed two paths that could leave a volume's used_sectors counter below the
number of sectors actually stored in it. Once the counter went negative the
background sector pruner panicked, which exited the daemon without writing
anything to the log and recurred on every restart.
Also fixed hostd recalculate skipping volumes that have no sectors left,
which meant it could report success without repairing a volume whose rows had
all been deleted by an interrupted removal.
Hosts already affected have their counters repaired automatically by a database
migration on the next startup.