replicate improvements

[yaroslav-gwit]

vm replicate will be moved into a separate command replicate (just like vm snapshot -> `snapshot). This needs to happen in order to support Jails replication as well as the VMs replication.

Here is the list of improvements I want to introduce in the process:

• multiple transports support, not only SSH (I'll start adding more transports later, but an initial support for multiple transports will be implemented within this issue)
• limited SSH users, with only access to the necessary ZFS functions (using ZFS delegation), the idea is to create limited usr_zfs_send and usr_zfs_recv users dedicated to ZFS replication only (principle of least permissions - even if the user account is compromised, the attacker will be limited to only what the replication user can do)
• support for Jail replication
• background replication (to avoid wrapping up the replication process inside the tmux, and to allow more flexible automation using our new custom scheduler)

And some things I will not look into implementing any time soon:

• support for regular ZFS datasets (eg specifying a dataset as the source, instead of the VM name or a Jail name)
• ability to specify the receiving dataset
• pull-based replication support (for purely backup systems, isolated from the rest of the cluster)

But all the above can be performed by the Sanoid/Syncoid combo, as Hoster isn't particularly picky as long as the VM disk is in the correct path. Hoster on the other hand has a totally different issue scope: replicate the VMs/Jails to another system with as low effort as possible, to provide the HA capabilities and a basic level of hyper-convergence.