cync_lan (Home Assistant custom_component) v2.6.3

[github-actions[bot]]

Fixes the hub query sensors ("Hub Firmware", "Hub Clock") polling the mesh far
harder than intended.

These are diagnostic entities, disabled by default, and their class docstring
described the poll interval as "deliberately long". It was never actually set:
the class asked Home Assistant to poll it and no SCAN_INTERVAL existed
anywhere in the platform, so they ran on HA's 30-second default.

That matters because each poll puts a real command on the mesh and then blocks
for up to 10 seconds waiting for a reply. This command family's transport is
unconfirmed (see docs/mesh_opcodes.md), so on hardware where it simply does
not answer, every poll cost a timeout warning plus Home Assistant's own
"Update of sensor ... is taking over 10 seconds" - measured on a real system at
around 5,700 log lines a day, from a sensor that had never once produced a
value.

They now run on their own 15-minute timer rather than HA's polling, so the
other sensors on the platform - all cheap local reads - stay responsive.
Neither firmware version nor hub clock drifts meaningfully in that window.

If you enabled either of these and saw nothing but timeouts, that behaviour
itself is unchanged: the command may not be answerable on your hub. It is just
quiet about it now. The matching core-library fix stops the repeated warning
and stops waiting out the full timeout when there was no connection to send
the request on in the first place - relevant right after a restart, when Cync
devices can take many minutes to reconnect.

---

This discussion was created from the release cync_lan (Home Assistant custom_component) v2.6.3.