The i2c-hid driver would quietly fail to probe the i2c-hid sensor-hub
with an ACPI device-id of SMO91D0 every other boot.

Specifically, the i2c_smbus_read_byte() "Make sure there is something at
this address" check would fail every other boot.

It seems that the BIOS does not properly reset/power-cycle the device
leaving it in a confused state where it refuses to respond to i2c-xfers.
On boots where probing the device failed, the driver-core puts the device
in D3 after the probe-failure, which causes the probe to succeed the next
boot.

Putting the device in D3 from the shutdown-handler fixes the sensors not
working every other boot.

This has been tested on both a Lenovo Miix 2-10 and a Dell Venue 8 Pro 5830
both of which use an i2c-hid sensor-hub with an ACPI id of SMO91D0.

Note that it is safe to call acpi_device_set_power() with a NULL pointer
as first argument, so on none ACPI enumerated devices this change is a
no-op.

Cc: Kai-Heng Feng <kai.heng.feng@canonical.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Acked-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>