Handle Wemo startup race condition.

[pavoni]

Description:
I've just started to see some errors on startup of WeMo switches.

This can result in not setting initial state - and the devices being unavailable.

Suspect this is a race condition where the callback from the wemo comes back before the HA component is initialized. I guess this could be because of other HA changes have changed the timing on startup.

This PR copes with getting initial state correct - even if the HA component isn't initialized.

Checklist:

• Local tests with tox run successfully.
• TravisCI does not fail. Your PR cannot be merged unless CI is green!
• Fork is up to date and was rebased on the dev branch before creating the PR.
• Commits have been squashed.
  script/gen_requirements_all.py.
  • New files were added to .coveragerc.
• If the code does not interact with devices:
  • Tests have been added to verify that the new code works.