Error in btrfs-receive when directory was replaced with subvolume between backups

One the directories in my machine stores videos that I download, which means there are lots of big files being put there, and roughly the same amount being deleted, very often. So this was filling-up my backup HD. I transformed it into a btrfs subvolume using commands similar to this:

```sh
mv $DIR $DIR.old
btrfs subvolume create $DIR
cp --reflink=always $DIR.old/. $DIR
rm -r $DIR.old
```

But now, when I run `btrbk`, `btrfs receive` throws an error. The relevant lines when running it with `--log debug` are these:

```
unlink o822875-724-0/$FILE
rmdir o822875-724-0
ERROR: rmdir o822875-724-0 failed: Directory not empty
```

I'm not sure this is the appropriate repo where to open this issue, but I thought I'd try.

The question: how do I resume backups? I don't care about preserving the contents of that directory across snapshots or even backups. I'd like to avoid doing a complete backup from scratch.

Start by reproducing the reported btrbk run with --log debug and examine the btrfs receive operation around the unlink and rmdir messages. Determine how backups can resume after a directory is replaced by a subvolume, and consider the work complete when incremental backups proceed without requiring a full backup from scratch.

Written by the indexing model from the issue text.