DeviceInstanceMapper: do not skip devices with inputDeviceError

I have a bunch of non-pushbuttons switches connected to some of my
Lunatone DALI-2 MC pushbutton couplers. These devices unfortunately
set the "inputDeviceError" bit whenever any of its inputs is
closed/shorted for longer than T_stuck.

The 62386-301:2017 says (Table 4, page 15) that this timer is up to
255s, which is definitely not enough for a door sensor. The same
standard then says that such errors should be propagated to bit 0 in the
"instanceErrorByte", which python-dali does not query. However, my
devices unfortunately also propagate this status into the
"inputDeviceError", and that means that the initial discovery of
instance numbers to instance types won't recognize these buttons at
all.

I was not able to find a reference in the standard whether any
"instanceError" propagates to the "inputDeviceError", but even if this
was an incorrect behavior, these devices exist, one can buy them, and
presumably using them as a kind-of persistent contact inputs still makes
sense; just to provide some data, I have:

- magnet and a reed switch for door status monitoring,
- a "permanent" switch for disabling automation at a hallway where our
  baby sometimes sleeps.

I think that both use cases make sense, and I would prefer python-dali
to not skip over some input devices which actually work just fine.

Fixes: 0cfddb7 Add support for DALI 24-bit event messages

dali/device/helpers.py

@@ -127,8 +127,7 @@ def autodiscover(
             if isinstance(rsp, QueryDeviceStatusResponse):
                 # Make sure the status is OK
                 if (
-                    rsp.input_device_error
-                    or rsp.short_address_is_mask
+                    rsp.short_address_is_mask
                     or rsp.reset_state
                 ):
                     continue

dali/tests/test_device_sequences.py

@@ -77,7 +77,7 @@ class DeviceFakeError(fakes.Device):
     _device_status = 0b00000001
 
 
-def test_device_autodiscover_skip_bad(fakes_bus):
+def test_device_autodiscover_dont_skip_bad(fakes_bus):
     dev_inst_map = DeviceInstanceTypeMapper()
     # Add an extra fake device, one with an error bit set
     fakes_bus.gear.append(
@@ -90,8 +90,8 @@ def test_device_autodiscover_skip_bad(fakes_bus):
     assert rsp.input_device_error
 
     fakes_bus.run_sequence(dev_inst_map.autodiscover())
-    # The device in error mode shouldn't have had its instances counted
-    assert len(dev_inst_map.mapping) == 12
+    # The mere fact that a device has "inputDeviceError" set does not mean it is not usable
+    assert len(dev_inst_map.mapping) == 16
 
 
 class DeviceNoInstances(fakes.Device):