v3.5.0

Device health and reliable backups

New

• Connectivity diagnostic binary sensor per device. on while the
  websocket is connected and the HTTP API answers a light probe
  (GET /rooms, 5 s timeout, every 60 s; /discovery on firmware without
  /rooms). It turns off as soon as either fails, and the entity itself
  stays available. Attributes: last_probe_ok, websocket_connected,
  ws_reconnects. Only transitions are logged (one WARNING when the device
  goes offline, one INFO when it comes back). No external Ping entry is
  needed any more to condition an automation on the device being reachable.
• Last backup diagnostic timestamp sensor, with a backup_count
  attribute. It is derived from the backup files already on disk, so nothing
  extra is written to the device or the host.
• Number of backups to keep option (1 to 50). Existing entries keep 5.

Changed

• espsomfy_rts_enhanced.backup (and the Backup button) now fails with an
  error when the backup could not be written, or immediately, without any
  request, when the Connectivity sensor is off. Automation traces show the
  failure instead of a success.
• Old backups are rotated only after a new one has been written.
• The Backup button no longer turns unavailable when the websocket drops:
  Home Assistant skips unavailable entities in service calls, which turned a
  backup of an offline device into a silent no-op.

Fixed

• Shade covers failed to be created after the firmware-room area seeding
  added since v3.4.0 ('ESPSomfyController' object has no attribute 'room_name'). This never shipped in a release.

No existing entity id or unique id changes; the new entities are additive.
See docs/Services.md
for a nightly backup automation conditioned on the Connectivity sensor.