etcd recovery on bootstrap

[smira]

Feature Request

Given that etcd backup was created via #3408, in case of control plane complete failure provide a way to recover a control plane from the backup.

Description

Failed control plane nodes are decommissioned.

New control plane is created with same machine configuration as before, it enters the waiting loop for etcd to join.

User can now do talosctl etcd bootstrap --from-backup=etcd.db to bootstrap etcd from the snapshot recovering the state of the cluster.

Tests

Take snapshot, wipe all control plane nodes, wait for the nodes to enter bootstrap waiting state, recover from the backup, cluster should be healthy.

[smira]

Implementation notes:

1. Add an API to upload (stream) snapshot to the Talos node to some predefined location, e.g. /var/lib/etcd.snapshot. (it should on EPHEMERAL partition to allow backup size larger than RAM size).
2. Add argument to bootstrap API to recover (boolean) which expects valid snapshot to be available in the predefined location.