Allow commiting after each backup is sent

I notice that after all my backups are `send`-ed (😃), and there is no `btrbk` process running, there is still *very* heavy IO on my target disk. This can go on for several minutes.

I suspect that this is why I'm not able to immediately unmount the backup target disk sometimes even though no non-kernel processes are accessing it.

With no `btrbk` running and no other usage of my backup disk, `iostat` reported 100% `util`. My chrome browser and another application stopped responding to keyboard or mouse input.

`btrfs-transaction` was the kernel process running if I recall correctly.

To enable a backup drive to be unmounted sooner after the last backup is sent, would you consider facilitating a:

```
btrfs filesystem sync /targetFS
```

After each backup is sent?

Perhaps an option like:

`backup_sync_after` with suggested values:

* `none` - current behaviour
* `each` - after each `send` completes
* `after` - after all backups have been sent (for parity with `btrfs_commit_delete` - it may speed up triggering the system generated transaction to explicitly request it - but just a guess)