Skip in-container sshd setup for single-node container runs

[atnair-amd]

Summary

ContainerOrchestrator.setup_sshd() ran a fixed command list ending in
/usr/sbin/sshd -p2224 and asserted every step exited 0 — for every
container run, regardless of node count (cvs/tests/conftest.py calls it
unconditionally). The in-container sshd exists solely so MPI (mpirun's
plm_rsh_args -p 2224, see BaremetalOrchestrator.get_mpi_command) can reach
peer ranks on other nodes. A single-node run execs directly via docker exec
and never distributes over MPI, so the sshd setup is dead weight on one host.

Consequences before this change:

• A single-node run on a minimal image without /usr/sbin/sshd failed the whole
  orch fixture (pytest.fail("Failed to setup sshd in container")) and never
  ran the workload — surfaced only as the generic SSH setup command failed.
• Single-node, no-SSH suites on sshd-equipped images still paid a pointless step
  that can fail on unrelated edge cases (e.g. empty ~/.ssh mount).

Change

• Guard setup_sshd() to return True early when len(self.hosts) <= 1,
  placed after the container_id precondition. Host count lives on the
  orchestrator, so the decision belongs there; the fixture stays unchanged and
  any future caller inherits the behavior.
• Multinode runs are unaffected — they still set up and validate sshd.

Tests

• cvs/core/orchestrators/unittests/test_container.py: single-node skips (no
  runtime.exec), container_id precondition still raises even single-node,
  multinode still attempts setup.

Out of scope

• Multinode + no-sshd image still fails (correctly — MPI needs sshd); its error
  message could be clearer, tracked separately.
• No new config key; behavior keys purely off node count.

Gate

make test (mirrors CI .github/workflows/ci.yml, Python 3.10): make fmt
clean (213 files unchanged), Ran 370 tests ... OK, 42 CLI tests passed.