refactor: SSHClient should not be a singleton

[mdonadoni]

After adding the Compute4PUNCH backend, the SSHClient is now used by two different backends. However, the client is implemented as a singleton, so it can't be used by the two backends at the same time. This means that workflows whose jobs run on multiple backends will not work if more than one of the chosen backends rely on SSH to dispatch jobs.

While this is not a big issue for now, as we do not expect to have instances that enable both Slurm@CERN and Compute4PUNCH, we should in any case refactor SSHClient to not be a singleton.