v0.5.1

Changed

• Upgraded quilt-hp-python dependency to >=0.5.1
  • QuiltClient.close() now clears the cached token, preventing stale token access after close
  • invalidate_snapshot() log level lowered from WARNING to DEBUG, reducing log noise
  • invoke_refresh_callback deduplicated into a single shared implementation; eliminates repeated inspect.signature() calls on every token-refresh event
  • FanSpeed.to_wire() / LouverAngle.to_wire() no longer re-allocate mapping dicts on every call

Fixed

• diagnostics.py accessed hass.data[DOMAIN][entry.entry_id] (old pattern) instead of
  entry.runtime_data; every call to the HA diagnostics page raised KeyError
• Config flow reconfigure + OTP: when the user changed email and OTP was required, the
  success path called _create_entry and created a duplicate entry instead of updating the
  existing one; added _reconfigure_entry tracking so the OTP success path routes to
  async_update_reload_and_abort in reconfigure context
• Config flow async_step_otp: FlowError (e.g. AbortFlow) was caught by the generic
  except Exception handler and converted to an "unknown" error, silently preventing reauth
  from completing when OTP was required; FlowError is now re-raised
• Energy refresh tasks are now created via config_entry.async_create_background_task so
  they are cancelled on entry unload rather than outliving it
• Stream-triggered energy fetch did not call async_set_updated_data after updating
  energy_by_space_id, so energy sensor entities would not re-render until the next
  unrelated stream push; a new _update_energy_and_notify() wrapper calls
  async_set_updated_data only when a fetch actually occurred (not on rate-limited early return)
• Schedule switch called coordinator.async_request_refresh() directly after writes instead
  of _async_refresh_if_not_streaming(), causing unnecessary polls while the gRPC stream
  is active
• Removed redundant targeted dict mutations in stream handlers that were immediately
  overwritten by async_set_updated_data
• Energy sensors stopped updating while the gRPC stream was active because
  async_set_updated_data cancels pending coordinator polls; the coordinator now
  explicitly checks whether an energy refresh is due on each stream push and fetches
  if so, ensuring reliable 30-minute energy updates regardless of stream activity
  (thanks @c00w, #3)