Add support for diskless machines to system monitor

[jrbourbeau]

psutil.disk_io_counters() returns None on diskless machines. This PR ensures we handle that case in our system monitor.

• Closes Dask scheduler initialisation fails when running on a node without "physical disk" #5254
• Tests added / passed
• Passes black distributed / flake8 distributed / isort distributed

[jrbourbeau]

There are a couple of unrelated test failures which I suspect are related to the recent changes in #5120

  File "/Users/runner/work/distributed/distributed/distributed/core.py", line 502, in handle_comm
    result = await result
  File "/Users/runner/work/distributed/distributed/distributed/scheduler.py", line 4327, in add_worker
    await self.handle_worker(comm=comm, worker=address)
  File "/Users/runner/work/distributed/distributed/distributed/scheduler.py", line 5445, in handle_worker
    await self.handle_stream(comm=comm, extra={"worker": worker})
  File "/Users/runner/work/distributed/distributed/distributed/core.py", line 575, in handle_stream
    handler(**merge(extra, msg))
  File "/Users/runner/work/distributed/distributed/distributed/scheduler.py", line 5320, in handle_task_finished
    r: tuple = self.stimulus_task_finished(key=key, worker=worker, **msg)
  File "/Users/runner/work/distributed/distributed/distributed/scheduler.py", line 4716, in stimulus_task_finished
    r: tuple = parent._transition(key, "memory", worker=worker, **kwargs)
  File "/Users/runner/work/distributed/distributed/distributed/scheduler.py", line 2248, in _transition
    for plugin in self.plugins.values():
RuntimeError: dictionary changed size during iteration

cc @douglasdavis