Move remote-folder creation to the FractalSSHSlurmexecutor

[tcompa]

(observed with @zonia3000)

When a job is marked as failed but the corresponding dataset has no history (because the error took place before running the first task), the error-displaying in the webclient can be improved - see fractal-analytics-platform/fractal-web#546.

Nevertheless, for the specific case that we hit this time, we can avoid the failed-job-with-no-dataset-history situation by implementing the suggestion in the FIXME in the following block

fractal-server/fractal_server/app/runner/v2/__init__.py

Lines 196 to 210 in 5b4be90

	elif FRACTAL_RUNNER_BACKEND == "slurm_ssh":
	WORKFLOW_DIR_REMOTE = (
	Path(settings.FRACTAL_SLURM_SSH_WORKING_BASE_DIR)
	/ WORKFLOW_DIR_LOCAL.name
	)
	# FIXME SSH: move mkdir to executor, likely within handshake
	fractal_ssh.mkdir(
	folder=str(WORKFLOW_DIR_REMOTE),
	)
	logger.info(f"Created {str(WORKFLOW_DIR_REMOTE)} via SSH.")
	else:
	logger.error(
	"Invalid FRACTAL_RUNNER_BACKEND="
	f"{settings.FRACTAL_RUNNER_BACKEND}."
	)